SPECIAL COMMUNICATION

Guest Authorship and Ghostwritingin Publications Related to Rofecoxib

ACaseStudyofIndustryDocumentsFromRofecoxibLitigation

Joseph S. Ross, MD, MHSKevin P. Hill, MD, MHSDavid S. Egilman, MD, MPHHarlan M. Krumholz, MD, SM

UTHORSHIP IN BIOMEDICAL

publication provides recog-nition while establishing ac-countabilityandresponsibil-ity.Guestauthorshiphasbeendefinedasthedesignationofanindividualwhodoesnotmeetauthorshipcriteriaasanauthor.

1,2

Itwasidentifiedin16%ofre-search articles, 26% of review articles,and 21% of editorials in a survey of 6peer-reviewedmedicaljournals,

inad-dition to 41% of Cochrane reviews.

Ghostwriting has been defined as thefailuretodesignateanindividual(asanauthor) who has made a substantialcontribution to the research or writ-ingofamanuscript.

Ghostwritingwasdemonstrated in 13% of research ar-ticles,10%ofreviewarticles,6%ofedi-torials, and 11% of Cochrane re-views

3,4

; other research has foundsimilar rates.

Two studies have characterized thepracticesofguestauthorshipandghost-writingusingindustrydocuments,oneexamining practices related to gaba-pentin by Pfizer Inc and Parke-Davis,Division of Warner-Lambert Com-pany,

theothersertralinebyPfizerInc.

However,thesestudieswerefocusedonhowtheresearchandpublicationstrat-

See also pp 1813 and 1833.

AuthorAffiliations:

DepartmentofGeriatricsandAdultDevelopment,MountSinaiSchoolofMedicine,NewYork,NewYork(DrRoss);DepartmentofPsychiatry,HarvardMedicalSchool,Boston,Massachusetts,andMcLeanHos-pital,Belmont,Massachusetts(DrHill);DepartmentofCommunityHealth,BrownUniversitySchoolofMedi-cine,Providence,RhodeIsland(DrEgilman);andRob-ertWoodJohnsonClinicalScholarsProgramandSectionof Cardiovascular Medicine, Department of Medicine,Section of Health Policy and Administration, School ofPublicHealth,YaleUniversitySchoolofMedicine,andCenterforOutcomesResearchandEvaluation,Yale-NewHavenHospital,NewHaven,Connecticut(DrKrumholz).

CorrespondingAuthor:

JosephS.Ross,MD,MHS,MountSinaiSchoolofMedicine,1GustaveL.LevyPl,Box1070,NewYork,NY10029 (joseph.ross@mssm.edu).

Context

Authorship in biomedical publication provides recognition and establishesaccountabilityandresponsibility.Recentlitigationrelatedtorofecoxibprovidedauniqueopportunity to examine guest authorship and ghostwriting, practices that have beensuspected in biomedical publication but for which there is little documentation.

Objective

To characterize different types and the extent of guest authorship andghostwriting in 1 case study.

Data Sources

Court documents originally obtained during litigation related to ro-fecoxibagainstMerck&CoInc.Documentswerecreatedpredominantlybetween1996and 2004. In addition, publicly available articles related to rofecoxib identified viaMEDLINE.

DataExtraction

Alldocumentswerereviewedbyoneauthor,withselectedreviewby coauthors, using an iterative process of review, discussion, and rereview of docu-ments to identify information related to guest authorship or ghostwriting.

Data Synthesis

Approximately 250 documents were relevant to our review. For thepublication of clinical trials, documents were found describing Merck employees work-ing either independently or in collaboration with medical publishing companies to pre-paremanuscriptsandsubsequentlyrecruitingexternal,academicallyaffiliatedinvestiga-torstobeauthors.Recruitedauthorswerefrequentlyplacedinthefirstandsecondpositionsoftheauthorshiplist.Forthepublicationofscientificreviewpapers,documentswerefounddescribingMerckmarketingemployeesdevelopingplansformanuscripts,contractingwithmedical publishing companies to ghostwrite manuscripts, and recruiting external, aca-demically affiliated investigators to be authors. Recruited authors were commonly thesole author on the manuscript and offered honoraria for their participation. Among 96relevant published articles, we found that 92% (22 of 24) of clinical trial articles pub-lished a disclosure of Merck’s financial support, but only 50% (36 of 72) of review ar-ticles published either a disclosure of Merck sponsorship or a disclosure of whether theauthor had received any financial compensation from the company.

Conclusions

Thiscase-studyreviewofindustrydocumentsdemonstratesthatclini-caltrialmanuscriptsrelatedtorofecoxibwereauthoredbysponsoremployeesbutof-ten attributed first authorship to academically affiliated investigators who did not al-ways disclose industry financial support. Review manuscripts were often prepared byunacknowledged authors and subsequently attributed authorship to academically af-filiated investigators who often did not disclose industry financial support.

JAMA. 2008;299(15):1800-1812

www.jama.com

1800

JAMA,

April 16, 2008—Vol 299, No. 15

(Reprinted)

on April 15, 2008www.jama.comDownloaded from

egy of the companies was used to pro-moteandmarkettheproducts.Nostud-ies have used internal documents tocharacterize the role of authorship incollaborationsbetweenindustryandthemedical profession.RecentlitigationagainstMerck&CoInc related to rofecoxib provided aunique opportunity to examine thepracticeofguestauthorshipandghost-writingrelatedtotheresearchandpro-motion of this medication. Our objec-tive was to provide a review using acase-study exploration of court docu-ments, in tandem with a review of themedicalliterature,todescribetheprac-ticeofguestauthorshipandghostwrit-ing related to rofecoxib.Documents used for this article areposted at http://dida.library.ucsf.edu.

METHODS

In the course of the combined trials of

Cona vs Merck and Co, Inc

(No. ATL-L-3553-05,NewJerseySuperiorCourt,Atlantic City) and

McDarby vs Merckand Co, Inc

(No.ATL-L-1296-05,New Jersey Superior Court, Atlantic City),millionsofdocumentsweremadeavail-abletoandarchivedinanintegratedda-tabase maintained by the plaintiff’s at-torneys.Thesedocumentswerecreatedbetween 1996 and 2004 and includedMerckinternalandexternalcorrespon-dence, reports, and presentations. Asconsultants to attorneys on the behalf ofplaintiffs,wehadcompleteaccesstoall archived documents. One investi-gator (J.S.R.) searched the database toextract a subset of documents relatedto authorship (

IGURE

). The searchwasperformedusingthedatabasekey-wordsearchfunctionandincludedthefollowingsearchterms:

clinicaltrial,au-thor,authorship,review,manuscript,

and

publication

, along with terms encom-passing the names of Merck scientists,thenamesofacademicallyaffiliatedau-thors of clinical trials, the names andnumbersofclinicaltrials,thenamesof medicalpublishingcompanies,andthenames of journals. The search identi-fied approximately 20000 documentsthat included 1 or more of the key-wordterms.Documentnumbersareanapproximation because informationwithinonedocumentmayoverlapwithanother, making it difficult to deter-minetheexactnumberofdistinctdocu-ments. For example, 1 document mayinclude a string of 2 e-mails, whereasanotherdocumentmayincludeastringof 5 e-mails, including the prior 2.Oneinvestigator(J.S.R.)searchedthedocumentsidentifiedusingtheauthor-shipkeywordstodetermineifeachwasactuallyrelatedtoordiscussedauthor-ship,examiningthedocumenttitlesandthe content within the database. “Re-lated to or discussed authorship” re-fersspecificallytoexaminationforau-thorship of manuscripts describingnonpharmacological, human partici-pantclinicaltrialresultsorscientificre-views(orjournalsupplements)thatin-cluded an external, academicallyaffiliated (non-Merck employee) au-thor. Approximately 250 documentswere identified, the majority of whichwere Merck internal correspondenceandpublicationreports,alongwithex-ternalcorrespondencebetweenMerckand medical publishing companies.Twoinvestigators(J.S.R.andK.P.H.)reviewed these 250 documents usingthe principles of grounded theory, aninductiveapproachinwhichsourcema-terialwasusedtogenerateideasratherthan to test a preestablished hypoth-esis.

This method has been applied tostudy issues at the intersection of liti-gationandhealth,particularlywithto-bacco,

9,10

andmorerecentlywithphar-maceutical

products.Wefirstreviewedthedocumentstoidentifybroadthemesreflecting the practice of ghostwritingand guest authorship. Next, pertinentdocuments were reviewed again by alloftheauthors,usinganegotiatedcon-sensusprocesstoreachourfinalinter-pretation.Thisprocessultimatelygen-erated a single agreed-upon set of themes,aswellasdocumentsandquo-tations to illustrate each theme.After determining themes, 2 inves-tigators (J.S.R. and K.P.H.) again re-viewed these 250 documents to iden-

Figure 1.

Document and Manuscript Identification Flowchart

96Distinct articles describing clinicaltrial results or scientific reviewsincluded in analysis

55Published articles identified in MEDLINE41Published articles identified in MEDLINE Approximately 20 000 documents identifiedusing keywords related to authorship All documents produced as part of litigationrelated to rofecoxib searched using afunction within the computerized databasefor keywords related to authorship Approximately 19

750 documentsexcluded that did not relate to anddid not discuss authorshipMEDLINE searched for other rofecoxib-related articles authored by academicallyaffiliated investigators that had beenidentified as first authors withindocuments Approximately 250 documents searched toidentify manuscripts describing clinicaltrial results or scientific reviews (including journal supplements) discussed internallywithin Merck prior to publication

Seethe“Methods”sectionfordetaileddescriptionsofthesearchtermsandthenumberofdocumentssearchedand for the definition of “related to or discussed authorship.”

Identification of these manuscripts does not imply that each was guest authored or ghostwritten; we exam-ined these manuscripts because we believed their discussion within internal documents (or the discussion ofspecific authors) suggested that Merck was aware of the publication and perhaps had provided support for the project.

GUEST AUTHORSHIP AND GHOSTWRITING IN ARTICLES ON ROFECOXIB

(Reprinted) JAMA,

April 16, 2008—Vol 299, No. 15

1801

on April 15, 2008www.jama.comDownloaded from

tify articles describing clinical trialresults or scientific reviews (includ-ingjournalsupplements)discussedin-ternally within Merck prior to publi-cation that proposed an external,academically affiliated investigator asan author. No documents were ex-cluded as part of this search; all docu-ments were related to 1 or more of theidentified manuscripts (most manu-scriptswerediscussedwithin



1docu-ment). The published articles weresubsequentlyidentifiedviaMEDLINE.This search was supplemented withMEDLINEqueriesforotherrofecoxib-related articles authored by academi-cally affiliated investigators identifiedas first authors within documents,whichwerefoundbysearchingfortheauthor’s name and “rofecoxib” or “cy-clooxygenase inhibitor.” This searchidentified96publishedarticles.Impor-tantly, identification of these articlesdoesnotimplythateachwasguestau-thored or ghostwritten; we examinedthesearticlesbecausewebelievedtheirdiscussion within internal documents(or the discussion of specific authors)suggestedthatMerckwasawareofthemanuscript prior to publication andperhaps had provided support for theproject.All published articles were catego-rized as to whether a manuscript’s co-author was affiliated with Merck (ie, aMerck employee), whether the pub-lishedarticleincludedanyfinancialdis-closure, whether the published articleincludedafinancialdisclosureofMercksupport,andwhetherthepublishedar-ticle included a financial disclosure of Merck support by at least 1 of the aca-demic authors. For articles in whichthere was neither a published disclo-sure of Merck financial support nor apublished disclosure of Merck finan-cial support by 1 of the academic au-thors,otherarticlesbytheacademicau-thors published within 2 years of therelevant article were examined to de-termineiftheyhaddisclosedMerckfi-nancial support.This research was deemed exemptfromnormalreviewbytheYaleUniver-sity Human Investigation Committee.

RESULTS

Review of internal documents and thepublishedliteraturerevealed3keyfind-ings related to guest authorship andghostwriting: the first focused on thepublication of clinical trials, the sec-ond focused on the publication of re-viewpapers,andthelastwasrelatedtofinancial support disclosures.

Clinical Trial Manuscripts

Whenpublishingtheirownclinicaltrials(designed,conducted,andsponsoredbyMerck),documentswerefounddescrib-ing Merck scientists often working toprepare manuscripts and subsequentlyrecruiting external, academically affili-ated investigators to collaborate on themanuscript as guest authors. For in-stance,trial078(arandomized,double-blindstudytoinvestigatewhetherrofe-coxibcoulddelaytheonsetofAlzheimerdisease in patients with mild cognitiveimpairment) was designed and con-ductedprincipallybyscientistsatMerck.

IGURE

showsthetitleandauthorlistboth from draft

and published

ver-sionsofthemanuscriptdescribingthetrial. Both the title and the authorshipwere modified to attribute authorshipto 3 academically affiliated investiga-tors (first, second, and third authors)onthepublishedarticle,inadditiontothe 8 Merck scientists who are attrib-uted authorship on both the draft andpublished versions of the manuscript(1Merckscientistisattributedauthor-shiponthedraftbutnotthefinalmanu-script). Of note, only 1 of the 3 aca-demically affiliated investigators whoare attributed authorship on the pub-lishedarticlewasacknowledgedinthedraft version as a participating inves-tigatorintherofecoxib078studygroup.In an internal e-mail discussing whereto publish trial 078 as the draft is cir-culated, one of the Merck scientistsstates, “I think you should be the firstauthorsinceyouhavedonevirtuallyallofthewriting.”

Althoughtherearemi-nor differences in language and orga-nizationbetweenthedraftandfinalver-sionsofthemanuscript(particularlyintheabstract,asopposedtothetext),theresults presented are almost identical,reinforcing that the trial itself and theanalyseswerecompletebeforetheaca-demically affiliated investigators wereinvolved in the manuscript.This same pattern occurred for themanuscript describing the Merck pro-tocol901studies,whichcomparedtheefficacy of rofecoxib and naproxen inAsian and European populations. Ane-mailwrittenonbehalfofmembersof Merck’s publication committee to aMerck scientist states that the Euro-pean study had been prepared as amanuscriptandthatadraftwassharedwith the European authors, in addi-tion to describing 2 Merck employeeswho will prepare the manuscript de-scribingtheAsianstudy.

However,thefinal publication describes both trialsin a single article and lists neither of them as authors.

Documents were found describingotherexamplesofMerckrecruitingex-ternal,academicallyaffiliatedinvestiga-tors to collaborate as guest authors onmanuscripts prepared by Merck scien-tists. The first author of the Assess-mentofDifferencesBetweenVioxxandNaproxen to Ascertain Gastrointesti-nal Tolerability and Effectiveness(ADVANTAGE) study

described to a

New York Times

reporter in 2005,“Merck designed the trial, paid for thetrial,ranthetrial . . . Merckcametomeafterthestudywascompletedandsaid,‘We want your help to work on the pa-per.’ The initial paper was written atMerck,andthenwassenttomefored-iting.”

Theacademicallyaffiliatedau-thors of the Vioxx GI Outcomes Re-search(VIGOR)study,

inresponsetoanexpressionofconcernbythe

NewEn-glandJournalofMedicine

makeapointofassertingthatnoMerckemployeeorrepresentativewasinvolvedinthedraft-ing of their response, but do not dis-cuss who drafted the manuscript andwithrespecttocardiovasculareventsal-ludetonotdevelopingtheanalysisplan,not having access to the data, and notperformingtheanalyses.