Global collaborative networks on meta-analyses of randomized trials

17 Pages·2014·4.19 MB·English

Catalá-Lópezetal.BMCMedicine2014,12:15
http://www.biomedcentral.com/1741-7015/12/15
RESEARCH ARTICLE Open Access
Global collaborative networks on meta-analyses of
randomized trials published in high impact factor
medical journals: a social network analysis
Ferrán Catalá-López1,2*, Adolfo Alonso-Arroyo3,4, Brian Hutton5, Rafael Aleixandre-Benavent3 and David Moher5
Abstract
Background: Researchcollaborationcontributestotheadvancementofknowledgebyexploitingtheresultsof
scientificeffortsmoreefficiently,buttheglobalpatternsofcollaborationonmeta-analysisareunknown.Thepurpose
ofthisresearchwastodescribeandcharacterizetheglobalcollaborativepatternsinmeta-analysesofrandomizedtrials
publishedinhighimpactfactormedicaljournalsoverthepastthreedecades.
Methods:Thiswasacross-sectional,socialnetworkanalysis.WesearchedPubMedforrelevantmeta-analysesof
randomizedtrialspublisheduptoDecember2012.Weselectedmeta-analyses(includingatleastrandomizedtrialsas
primaryevidencesource)publishedinthetopsevenhighimpactfactorgeneralmedicaljournals(accordingtoJournal
CitationReports2011):TheNewEnglandJournalofMedicine,TheLancet,theBMJ,JAMA,AnnalsofInternalMedicine,
ArchivesofInternalMedicine(nowrenamedJAMAInternalMedicine),andPLoSMedicine.Opinionarticles,conceptual
papers,narrativereviews,reviewswithoutmeta-analysis,reviewsofreviews,andotherstudydesignswereexcluded.
Results:Overall,weincluded736meta-analyses,inwhich3,178authors,891institutions,and51countriesparticipated.
TheBMJwasthejournalthatpublishedthegreatestnumberofarticles(39%),followedbyTheLancet(18%),JAMA
(15%)andtheArchivesofInternalMedicine(15%).TheUSA,theUK,andCanadaheadedtheabsoluteglobalproductivity
rankinginnumberofpapers.The64authorsandthe39institutionswiththehighestpublicationrateswereidentified.
Wealsofound82clustersofauthors(onegroupwith55membersandonegroupwith54members)and19clusters
ofinstitutions(onemajorgroupwith76members).ThemostprolificauthorsweremainlyaffiliatedwiththeUniversity
ofOxford(UK),McMasterUniversity(Canada),andtheUniversityofBern(Switzerland).
Conclusions:Ouranalysisidentifiednetworksofauthors,institutionsandcountriespublishingmeta-analysesof
randomizedtrialsinhighimpactmedicaljournals.Thisvaluableinformationmaybeusedtostrengthenscientific
capacityforcollaborationandtohelptopromoteaglobalagendaforfutureresearchofexcellence.
Keywords:Authorship,Evidence-basedmedicine,Meta-analysis,Randomizedcontrolledtrial,Scientific
collaboration,Socialnetworkanalysis
Background evidenceinaresearchquestion.Whenperformedwelland
The past decades have seen the establishment of evidence reported completely, incorporating explicit and detailed
synthesis,particularlysystematicreviewsandmeta-analyses, methods and results, such studies produce information
as a key component of evidence based medicine (EBM) that can have undoubtedly major, immediate effects on
[1,2]. Meta-analyses of randomized trials have become medical practice, research agendas and the establishment
morewidelyacceptedbyclinicians,researchersandpolicy ofhealthcarepolicies.
makers as a useful tool to critically assess the totality of Importantmilestonesthatmayhaveencouragedresearch
inthisfield,fromthepointofviewofscientificpublications
*Correspondence:[email protected] and institutional development of EBM [1-4], include the
1DivisionofPharmacoepidemiologyandPharmacovigilance,Spanish
creationofinternationalresearchgroups,centers,andcon-
MedicinesandHealthcareProductsAgency(AEMPS),Madrid,Spain
2FundaciónInstitutodeInvestigaciónenServiciosdeSalud,Valencia,Spain sortia(suchastheCentreforEvidenceBasedMedicineand
Fulllistofauthorinformationisavailableattheendofthearticle
©2014Catalá-Lópezetal.;licenseeBioMedCentralLtd.ThisisanOpenAccessarticledistributedunderthetermsofthe
CreativeCommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),whichpermitsunrestricteduse,
distribution,andreproductioninanymedium,providedtheoriginalworkisproperlycredited.TheCreativeCommonsPublic
DomainDedicationwaiver(http://creativecommons.org/publicdomain/zero/1.0/)appliestothedatamadeavailableinthis
article,unlessotherwisestated.
Catalá-Lópezetal.BMCMedicine2014,12:15 Page2of17
http://www.biomedcentral.com/1741-7015/12/15
The Cochrane Collaboration in the 1990s) in addition American Medical Association (JAMA), Annals of Internal
to groups developing reporting guidelines to ensure arti- Medicine, Archives of Internal Medicine (now renamed
cles contain all essential information, such as QUOROM JAMAInternalMedicine)andPLoSMedicine.Specifically,
(Quality of Reporting of Meta-analyses) [5] and, more re- thefollowing termswere usedforPubMed:(‘meta-analy-
cently,PRISMA(PreferredReportingItemsforSystematic sis’[PublicationType]OR‘meta-analysisastopic’[MeSH
ReviewsandMeta-Analyses)[6]. Terms]OR‘meta-analysis’[AllFields])AND(‘randomized
Globalhealth challenges require research collaboration controlled trial’[Publication Type] OR ‘randomized con-
and multi-lateral programs on a global scale, owing to trolled trials as topic’[MeSH Terms] OR ‘randomized con-
the nature and magnitude of the public health problems. trolled trial’[All Fields]) AND (‘Lancet’[Journal] OR ‘N Engl
Noteworthyexamplesincludemajorenvironmental,polit- JMed’[Journal]OR‘JAMA’[Journal]OR‘BrMedJ’[Journal]
ical,andsocialdeterminantsof health,aswellas complex OR‘BrMedJ(ClinResEd)’[Journal]OR‘BMJ’[Journal]OR
and changing clinical issues related to the conditions and ‘PLoS Med’[Journal] OR ‘Ann Intern Med’[Journal] OR
riskfactorsthatcausethehighestburdenofdiseasearound ‘Arch Intern Med’[Journal]) AND (hasabstract[text] AND
the world [7-9]. Despite continuous efforts of individual ‘humans’[MeSHTerms]).Wealsoperformedcomplemen-
scientistsandinstitutionstoremedydeficienciesinhealth- tary hand-searches and reviewed references of identified
careeffectivenessand safety, multiple gapsand disparities eligiblereportstoidentifyadditionalmeta-analyses.
remain. Research collaboration contributes to the advance- We included two types of articles from the eligible
ment of knowledge by exploiting the results of scientific journals: original research reports and reviews (both in-
effortsmoreefficiently,buttheglobalpatternsofcollabor- corporating meta-analyses of randomized trials). Editorials,
ation on meta-analysis are unknown. Given that meta- commentaries, and other opinion articles were excluded.
analysescanprovidehigh-qualityclinicalevidenceregarding We also excluded conceptual papers, literature (narrative)
the robustness of the effects of healthcare interventions reviews, reviews of reviews, meta-analysis of observational
to inform medical practice, there is an urgent need to studies not considering randomized trials, single random-
evaluate and promote scientific activityand growth in the ized trials, and other study designs (such as cost-effective-
fieldofEBM[4,10-12]. nessanalysesandepidemiologicalstudies).
Social network analysis [13], the study of structure For the purposes of this study, we selected all articles
derived fromtheregularitiesin the patterningof relation- published in English and indexed in PubMed between
ships between social entities (which might be people or January1985andDecember2012.Oneresearcherwith
organizations),isgroundedintheassessmentofempirical expertise in evidence synthesis (FC-L) screened the titles
data, and can provide an appropriate approach to identify andabstracts,andidentifiedallpotentiallyeligiblearticles.
top scientists and researchers, groups of excellence, and Thesameresearcherexcludedthearticlesnotmeetingthe
leading institutions. It also offers information to assess pre-specifiedcriteria.
the citation patterns among papers within a specialty
[14], to identify gaps in the evidence from scientific re- Dataextraction
search [15], and to understand the structure and nature of For each included paper, we extracted information on
relationshipsandinteractionswithinascientificcommunity the year of publication, the journal title, and the authors’
that collaborate to better achieve common or compatible names, institutional affiliation(s), and country of origin.
goals[13,16]. This information was downloaded online through the
This study aimed to describe and characterize global Science Citation Index-Expanded (SCI-E) Web of Know-
collaborative patterns with regard to the conduct of meta- ledgeplatformversion5,inApril2013.TheWebofKnow-
analyses of randomized trials published over the past three ledge platform is a database that contains all the above
decadesinhighimpactfactormedicaljournals,byapplying information, including the full addresses of all authors of
techniquesfromsocialnetworkanalysis. every paper. We also used the SCI-E to determine the ex-
tent to which each study had been cited in the scientific
Methods peer-review literature using the ‘times cited’ number (that
Designandsample is, the number of times a publication has been cited by
In December 2012, we searched for reports of meta- other publications). A process of standardization was con-
analysesofrandomizedtrialsthatwereindexedinPubMed ductedtobringtogetherthedifferentnamesofaparticular
and published in one of the top seven high impact general author or institution. Specifically, one researcher (AA-A)
medical journals, asidentified in 2011, based onanimpact checkedthenamesbywhichanindividualauthorappeared
factorofatleast10(subjectcategories‘Medicine,General& in two or more different forms (for example, ‘Gordon
Internal’ of Journal Citation Reports, Thomson Reuters): Guyatt’or‘GordonHGuyatt’),usingcoincidenceinthat
The New England Journal of Medicine (NEJM),The Lancet, author’s place(s) of work as the basic criterion for nor-
the British Medical Journal (BMJ), the Journal of the malization(forexample,McMasterUniversity,Canada).
Catalá-Lópezetal.BMCMedicine2014,12:15 Page3of17
http://www.biomedcentral.com/1741-7015/12/15
Inthecaseofinstitutions,weunifiedthedifferentvariants applied an a posteriori threshold of three or more papers
tomatch the name recordedin publicdirectoriesofinsti- signed in co-authorship. The productivity and patterns of
tutions. Similarly, given that institutional names in many collaboration by author, institution and country were
records included two or more institutions (for example, analyzed.
university hospitals, research centers and academic institu- We used PAJEK [17], a software package for large
tions),weproceededtodistinguishbetweenthesenamesby network analysis that is free for non-commercial use,
recording all variations of any individual macroinstitution toanalyzeindicatorsandconstructsocialnetworks.
ascouldbeidentifiedforeachbibliographicrecord(forex-
ample,fortheinstitutionaladdress‘ReproductiveMedicine Results
Unit,DepartmentofObstetrics&Gynaecology,University Numberofmeta-analyses
of Adelaide, Queen Elizabeth Hospital, Australia’, the The PubMed search generated 804 records. Following
standardization approach was to present ‘University of screeningofabstractsandfulltextarticles,724publications
Adelaide, Australia’ separately from ‘Queen Elizabeth were retained, and 12 additional publications were added
Hospital,Australia’).Withallthisinformation,wecon- from complementary searches of reference lists, thereby
structedaMicrosoftAccessdatabase. yieldingafinalsampleof736includedmeta-analyses.The
processofstudyselectionispresentedinFigure1.
Dataanalysis
In this paper, we use the term ‘co-authorship’ to refer to Generalcharacteristicsofthesample
joint authorshipof a scientific paper byat leasttwo indi- TheBMJpublished thelargest number ofarticles(n=289;
viduals, and the term ‘institutional collaboration’ to refer 39%), followed by The Lancet (n=132; 18%), JAMA (n=
to joint authorship by different institutions. ‘Intensity of 113; 15%) and the Archives of Internal Medicine (n=112;
collaboration or threshold’ refers to the number used to 15%).Overall,736meta-analysesreceived130,644citations,
form clusters of authors and institutions (that is, the fre- of which 37,930 citations (29%) corresponded to BMJ, 34
quency of co-authorship between pairs of authors or of 911(27%) to The Lancet and 25,273 (19%) to JAMA. The
collaborationbetweeninstitutions),andreflectsacriterion number of publications increased exponentially over the
to label identifiable clusters as research groups. Collabor- study period (Table 1). Approximately three-quarters of
ation between authors (or institutions) was portrayed by the meta-analyses were reported during the most recent
calculating the number of papers, names, signatures and decade.
collaborations,theindexofsignaturesperpaperorcollab- The included meta-analyses had a median of 5 authors,
oration index (which is the mean number of signatures although 41 (6%) were single authored (Table 1). More
per paper), and the index of authors per paper (mean than a quarter (214 [29%]) of the first authors were
numberofauthorsperpaper,consideringonlythedifferent from the USA, with three countries (the USA, the UK
authors). A summary box with definitions of each of and Canada) accounting for more than two-thirds of
the measurements of collaboration is provided in the the meta-analyses published during the period of analysis
supplementary material (see Additional file 1: ‘Definitions (Table1).
ofcollaborativemeasurements’).
To construct co-authorship networks, we identified all
combinations of pairs of authors for each paper. The
number of co-authorships for each paper is related to
thenumberofauthorsasisequalto
m!
;
ðm−nÞ!n!
where m is the number of individual authors and n the
number of elements in the groups constructed. Once
co-authorship was quantified, we further established an
a posteriori threshold of two or more collaborations be-
tween pairs of authors, inorder to reduce the number of
nodes and links that would prevent a clear view of the
network,andthuscentertheanalysisonthemoreintense
co-authorship relationships. The same approach was ap-
plied to institutional and country authorship to construct
Figure1Selectionofpublications.Flowchart.
the network of collaborations, although in this case, we
Catalá-Lópezetal.BMCMedicine2014,12:15 Page4of17
http://www.biomedcentral.com/1741-7015/12/15
Table1Generalcharacteristicsofthesampleofstudy Productionandcollaborationpatterns
Characteristic Number(%) Overall, 3,178 authors, 891 institutions and 51 countries
worldwide were involved in the sample of articles. We
Totalnumberofarticles 736(100)
identified 64 authors who published 5 or more papers
Articlesperjournala
(Table2).ThemostprolificauthorswereLau(15papers),
BMJ 289(39.3)
Guyatt(14),Peto(13),Yusuf(12),Cook(11)andJüni(11).
TheLancet 132(17.9) Manyofthemostprolificauthorsareaffiliatedwithonlya
JAMA 113(15.3) few academic institutions and/or medical centers; six are
ArchInternMed 112(15.2) affiliated with the University of Oxford (Peto, Collins,
Clarke, Baigent, Gray and Rothwell), five are affiliated
AnnInternMed 70(9.5)
with McMaster University (Guyatt,Yusuf, Cook, Douketis,
NewEnglJMed 13(1.8)
andEikelboom)andfiveareaffiliatedwiththeUniversityof
PLoSMed 7(0.9) Bern and/or the Inselspital – Bern University Hospital
Citationsperjournalb (Jüni, Trelle, Egger, Reichenbach and Nüesch). Apply-
BMJ 37,930(29.0) ing a threshold of 2 or more papers published as co-
TheLancet 34,911(26.7) authors (Figures 2, 3, 4, 5, 6, 7), we identified 82 clusters
of authors. Of these, 12 were identified as major co-
JAMA 25,273(19.3)
authorshipgroups(1with55members,1with54mem-
AnnInternMed 13,381(10.2)
bers,1with27members,1with15,4with14members,
ArchInternMed 11,945(9.1)
3 with 11 members, 1 with 10 members, 1 with 8 mem-
NewEnglJMed 6,992(5.3) bers, 5 with 7 members, 5with 6membersand 3with5
PLoSMed 212(0.1) members).
Articlesperperiodofpublication Institutional productivity was headed by McMaster
University (49 papers), University of Oxford (48 papers)
1985to1989 3(0.4)
and Harvard University (36 papers) (Table 3). Next came
1990to1994 53(7.2)
some of their affiliated hospitals or medical centers
1995to1999 140(19.0) (Brigham and Women’s Hospital and the Radcliffe In-
2000to2004 166(22.6) firmary, with 32 and 30, respectively). Applying a col-
2005to2009 231(31.4) laboration threshold of at least 3 papers signed with
2010to2012 143(19.4) inter-institutional collaboration, we identified 19 clusters
Countryofpublication(firstauthor)c comprisingatotalof120institutions(Figures8and9).Of
these, the most important institutional cluster comprised
USA 214(29.1)
76members.
UK 207(28.1)
The productivity ranking for countries with respect to
Canada 94(12.8) the number of papers (Table 4) was headed by the USA
Australia 36(4.9) (310papers),theUK(297papers)andCanada(143papers).
Switzerland 26(3.5) After these countries came Australia (70 papers), and Italy
and the Netherlands (57 papers each). The USA and the
France 26(3.5)
UKalsoheadedthelistofthenumberofdifferentcountries
Denmark 20(2.7)
with which they had collaborated, as well as the total
Italy 20(2.7)
number of collaborations. Figure 10 shows a visual repre-
Numberofauthorsperpaper sentation of the collaborative network between countries,
1 41(5.6) inwhichwecanseetherelationshipsofsomewithrespect
2to3 166(22.5) to others and the position that each occupies in the
networkasawhole.
4to6 297(40.3)
The 75 most cited articles by number of citations are
>7 232(31.5)
listedinthesupplementarymaterial(seeAdditionalfile2:
abPInLfOorSmMaetidoincinreefehrasstobethenetpoutballischiteadtiowneserkelcyeoivnelidnebysinalcleth1e9pOacpteorbsepru2b0l0is4h.ed ‘List of most cited meta-analyses’). Heavily cited meta-
inagivenjournal(forexample,289paperspublishedinBMJreceivedatotal analyses include randomized trials examining the health
of37,930citations).
effectsofpharmacologicalinterventionsin cardiologyand
cInformationwasincompleteforsomerecords.Weusedcountrydataofthe
firstauthorin704outof736papers,whilefor32papersthisinformationwas oncology (for example, antithrombotic trials, anti-platelet
retrievedfromthecorrespondenceaddress. trials, antihypertensive trials, lipid-lowering trials, chemo-
therapy for diverse cancers such as breast cancer, lung
cancer,orheadandneckcancers).
Catalá-Lópezetal.BMCMedicine2014,12:15 Page5of17
http://www.biomedcentral.com/1741-7015/12/15
Table2Rankingofmostprolificauthors(fiveormorepapers)andtheircollaborativepatterns
Author Primaryaffiliationat Yearoffirst Total Signatures,n Collaborations,n Maincollaborators
thetimeofpublication eligiblepaper papers,n (numberofpapers)
Lau,Joseph BrownUniversity,USAa 1992 15 90 54 ChalmersTCandIoannidis
JPA(5)
Guyatt,GordonH McMasterUniversity, 1996 14 104 76 CookDJ(5)
Canada
Peto,Richard UniversityofOxford,UK 1991 13 124 75 CollinsR(8)
Yusuf,Salim McMasterUniversity, 1989 12 83 54 CollinsR,EikelboomJW,Mehta
Canada SRandPogueJ(3)
Cook,DeborahJ McMasterUniversity, 1995 11 71 49 GuyattGH(5)
Canada
Jüni,Peter UniversityofBern, 1996 11 124 59 ReichenbachS(8)
Switzerland
Collins,Rory UniversityofOxford,UK 1991 9 78 45 PetoR(8)
Trelle,Sven UniversityofBern, 2007 9 103 67 JüniPandReichenbachS(7)
Switzerland
Bucher,HeinerC UniversityHospital 1996 8 60 41 GuyattGH(4)
Basel,Switzerland
Egger,Mathias UniversityofBern, 2001 8 65 49 JüniP,ReichenbachSand
Switzerland TrelleS(3)
Gluud,Christian CopenhagenUniversity 2001 8 40 21 WetterslevJ(3)
Hospital,Denmark
Gøtzsche,PeterC TheNordicCochrane 1995 8 27 17 JohansenHK(3)
Centre,Denmark
Jackson,JeffreyL ZablockiVAMedical 1997 8 33 23 BrowningRandO’Malley
Center,USA PatrickG(2)
Law,MalcolmR QueenMaryUniversity 1991 8 24 7 WaldNJ(8)
ofLondon,UK
Reichenbach,Stephan UniversityofBern, 2004 8 81 44 JüniP(8)
Switzerland
Sutton,AlexJ UniversityofLeicester,UK 2003 8 40 28 CooperNJ(3)
Suttorp,MaartenJ StAntoniusHospital, 2003 8 120 61 MaglioneM,MojicaWA,Morton
theNetherlands SCandShekellePG(5)
Wald,NicholasJ QueenMaryUniversity 1991 8 24 7 LawMR(8)
ofLondon,UK
Clarke,Michael UniversityofOxford,UK 1994 7 83 67 PetoR(4)
Furberg,CurtD WakeForestUniversity 1989 7 36 20 LokeYK,PsatyBMand
SchoolofMedicine,USA SinghS(3)
Baigent,Colin UniversityofOxford,UK 1996 6 50 26 CollinsRandPetoR(5)
Boersma,Eric ErasmusMedicalCenter, 2001 6 74 63 CaliffRM,SerruysP,SimesJ,
theNetherlands SimoonsMLandTopolEJ(2)
Boissel,Jean-Pierre UniversitéClaudeBernard 1992 6 42 26 FagardRHandGueyffierF(3)
Lyon1,France
Chalmers,ThomasC TuftsMedicalCenter, 1990 6 31 16 LauJ(5)
TuftsUniversity,USA
Douketis,JamesD McMasterUniversity, 2000 6 32 25 CrowtherMA(2)
Canada
Ebrahim,Shah LondonSchoolof 1997 6 70 60 SmithGD(4)
HygieneandTropical
Medicine,UK
Ioannidis,JohnPA StanfordUniversity,USA 1995 6 32 17 LauJ(5)
Khan,KhalidS QueenMaryUniversity 1996 6 39 27 BhattacharyaS,ChampaneriaR,
ofLondon,UK CooperK,JollyKandMiddleton
LJ(2)
Catalá-Lópezetal.BMCMedicine2014,12:15 Page6of17
http://www.biomedcentral.com/1741-7015/12/15
Table2Rankingofmostprolificauthors(fiveormorepapers)andtheircollaborativepatterns(Continued)
Klassen,TerryP UniversityofManitoba, 1996 6 48 36 MoherD(3)
Canada
Loke,YoonK UniversityofEastAnglia, 2000 6 18 7 SinghS(4)
UK
McAlister,FinlayA UniversityofAlberta, 2001 6 30 23 ArmstrongPW(2)
Canada
Moher,David OttawaHospitalResearch 1996 6 52 42 KlassenTP(3)
Institute,Canada
Roberts,Ian LondonSchoolof 1996 6 18 12 Allco-authorship(1)
HygieneandTropical
Medicine,UK
Sattar,Naveed UniversityofGlasgow, 2009 6 91 58 RayKKandSeshasaiSRK(5)
UK
Shekelle,PaulG RandCorporation,USA 2003 6 57 29 MaglioneM,MojicaWA,Morton
SCandSuttorpMJ(5)
Stone,GreggW ColumbiaUniversity 2005 6 120 77 LeonMB(4)
MedicalCenter,USA
Tognoni,Gianni ConsorzioMarioNegri 1993 6 78 69 MarchioliR,MarfisiRMand
Sud,Italy RoncaglioniMC(2)
Topol,EricJ ScrippsTranslational 1991 6 44 33 BoersmaE,CaliffRM,Simoons
ScienceInstitute,USA ML,TchengJEandVande
WerfF(2)
BischoffFerrari,HeikeA UniversityHospital 2004 5 48 26 DawsonHughesB,Staehelin
Zurich,Switzerland HB,WillettWCandWong
JB(4)
Briel,Matthias UniversityHospital 2006 5 51 43 BucherHC(3)
Basel,Switzerland
Eikelboom,JohnW McMasterUniversity, 2000 5 19 10 YusufS(3)
Canada
Ernst,Edzard UniversityofExeter,UK 1995 5 18 11 WhiteAR(3)
Fagard,RobertH UniversityofLeuven, 1997 5 51 36 BoisselJPandGueyffierF(3)
Belgium
Fergusson,Dean OttawaHospitalResearch 1998 5 33 22 HuttonB(3)
Institute,Canada
Glasziou,Paul BondUniversity,Australia 1993 5 26 21 Allco-authorship(1)
Godwin,Jon GlasgowCaledonian 1995 5 52 38 PetoR(4)
University,UK
Gray,Richard UniversityofOxford,UK 2001 5 48 36 ClarkeMandPetoR(3)
Hennekens,CharlesH FloridaAtlanticUniversity, 1995 5 28 19 Hebert,PR(4)
USA
Leibovici,Leonard RabinMedicalCenter, 2003 5 22 12 PaulM(5)
Israel
Maglione,Margaret RandCorporation,USA 2003 5 49 22 MojicaWA,MortonSC,Shekelle
PGandSuttorpMJ(5)
Mojica,WalterA RandCorporation,USA 2003 5 49 22 MaglioneM,MortonSC,Shekelle
PG,andSuttorpMJ(5)
Morton,SallyC UniversityofPittsburgh, 2003 5 49 22 MaglioneM,MojicaWA,Shekelle
USA PGandSuttorpMJ(5)
Nüesch,Eveline UniversityofBern, 2009 5 35 16 JüniP(5)
Switzerland
Paul,Mical RabinMedicalCenter, 2003 5 22 12 LeiboviciL(5)
BeilinsonHospital,Israel
Pignon,Jean-Pierre InstitutGustave-Roussy, 1999 5 53 46 BourhisJandMichielsS(2)
France
Catalá-Lópezetal.BMCMedicine2014,12:15 Page7of17
http://www.biomedcentral.com/1741-7015/12/15
Table2Rankingofmostprolificauthors(fiveormorepapers)andtheircollaborativepatterns(Continued)
Pocock,StuartJ LondonSchoolof 1995 5 67 54 BoisselJP,BoutitieF,FagardRH,
HygieneandTropical GueyffierF,HammCW,HuebWA,
Medicine,UK KingSBandRodríguezA(2)
Ray,KausikK StGeorge’sUniversity 2009 5 73 49 SaltarNandSeshasaiSRK(5)
ofLondon,UK
Rothwell,PeterM UniversityofOxford,UK 2003 5 36 19 BelchJFFandMeadeTW(3)
Seshasai,SreenivasaRK UniversityofCambridge, 2009 5 73 49 RayKKandSN(5)
UK
Simes,John NationalHealthand 2002 5 53 36 BaigentC,BlackwellL,CollinsR,
MedicalResearchCouncil KeechAandPetoR(3)
ClinicalTrialsCentre,
Australia
Smith,GeorgeDavey UniversityofBristol,UK 1993 5 63 54 EbrahimS(4)
Torgerson,DavidJ UniversityofYork,UK 1999 5 46 39 AdamsonSJandBellSyer
SEM(2)
Wetterslev,Jorn CopenhagenUniversity 2007 5 28 15 GluudC(3)
Hospital,Denmark
Wilt,TimothyJ MinneapolisVACenter 1998 5 40 32 MacDonaldR(4)
forChronicDisease
OutcomesResearch,
USA
aJosephLauwasbasedatTuftsUniversityMedicalCenterinBoston(USA)until2012,andtheinstitutionalnetworksrepresentthisfact.
Discussion important and relevant research questions that require a
Generally,thevisibilityandrecognitionofscientificre- complexconstructionofmulti-disciplinaryteamsofscien-
search activities benefits from increasing collaborative tistsandresearchers,large-scalescientificstructures,bud-
research [18]. Research collaboration plays an important getsofanunprecedentedscale,andwidespreadsharingof
role in science, policy and medicine [19,20]. Research scientific knowledge and data. Thus, meta-analysis can be
collaboration in the ‘Big Science’ era involves addressing considered a good example of ‘Big Science’ in medicine
Figure2Co-authorshipnetworks.Mostproductiveclusterofauthors,applyingathresholdoftwoormorepaperssignedinco-authorship.
Catalá-Lópezetal.BMCMedicine2014,12:15 Page8of17
http://www.biomedcentral.com/1741-7015/12/15
Figure3Co-authorshipnetworks.Secondmostproductiveclusterofauthors,applyingathresholdoftwoormorepaperssignedinco-authorship.
and clinical epidemiology [21], as quantitative evidence Therefore,promotingresearchcollaborationinevidence
synthesis is the application, in practice, of the principle synthesisisabletostrengthenresearchactivity,productiv-
that science is cumulative [22,23]. An obvious manifest- ityandimpact.
ationofthisistheobservedtrendof75newrandomized In general, we found a strong clustering of papers
trials and 11 new systematic reviews being published published in two British journals (BMJ and The Lancet
daily, with aplateauin this growth not yet reached [24]. accounted for 57% of meta-analyses), in contrast to other
Figure4Co-authorshipnetworks.Mainclustersofauthors(≥15members),applyingathresholdoftwoormorepaperssignedinco-authorship.
Catalá-Lópezetal.BMCMedicine2014,12:15 Page9of17
http://www.biomedcentral.com/1741-7015/12/15
Figure5Co-authorshipnetworks.Mainclustersofauthors(≤14members),applyingathresholdoftwoormorepaperssignedinco-authorship.
general medical journals (for example, the NEJM repre- and/or promoting the publication of high-quality quantita-
sentedlessthan2%).Wehypothesizedthatthesedifferent tiveevidencesynthesis.
findings between journals may potentially reflect an edi- Perhapsarelevantfindingisthatcollaborativenetworks
torial policy and/or preference, with the BMJ,The Lancet are expanding in multiple regions, revealing a discernable
and JAMA journals specifically being more interested in and well-established scientific community, with the most
Figure6Co-authorshipnetworks.Mainclustersofauthors(≤11members),applyingathresholdoftwoormorepaperssignedinco-authorship.
Catalá-Lópezetal.BMCMedicine2014,12:15 Page10of17
http://www.biomedcentral.com/1741-7015/12/15
Figure7Co-authorshipnetworks.Mainclustersofauthors(≤6members),applyingathresholdoftwoormorepaperssignedinco-authorship.
prolific authors and institutions having an important our research is definitely different from that paper, those
number of collaborations. As might be expected, the authors also showed that the USA, the UK and Canada
scientificcommunitycapturedbythenetworksiscentered havetakenleadershipintheproductionofcitationpapers,
on a nucleus of scientists and researchers from academia, but no first author from low or middle income countries
medical centers and health research institutes from ledoneofthemostcitedpapers.
western high-income countries (North America, Western The maps of scientific partnership show that authors
Europe and Australia/Oceania). Specifically, the most in- who are ‘leaders’ and thus who may contribute collab-
tense global collaborations took place between authors oration, have more frequent and intense collaboration
andinstitutionsfromtheUSA,theUKandCanada.How- between other authors and institutions from different
ever,althoughthesethreecountriesleadinthenumberof countries.Thestudyalsoidentifieshighlycohesivecluster
published high impact meta-analyses, the efforts during networks and provides considerable information on the
the period of study were global, with publications from structure that can be put to various purposes, such as
authors and institutions in more than 50 different coun- funding agencies designing strategies for future scien-
tries. Cultural links may have historically benefited some tific collaboration, agencies such as the World Health
countries through alliances with nations and regions that Organization promoting a global coordinated agenda for
speak the same language (as may be the case for the UK perceived high priority clinical topics, and sharing of reli-
throughallianceswithCommonwealthcountriesthatspeak able and innovative methodologies that can be linked to
English) and have adopted similar scientific and research world-classeducationalandtrainingopportunities.
structures [25]. However, there is a clear over-representa- Thereare severalpossible explanationsfor ourfindings.
tion of scientists based in western high-income countries, Theuseofmoderncommunicationandinformationtech-
and the limited participation of low and middle income- nologies, especially the Internet, has diminished the role
based researchers could warrant further pragmatic action. of geographical and territorial boundaries in the access
Giventhatresearchresourcesandfundingareoftenre- and transmissibility of information [27]. This has enabled
stricted,itistheresponsibilityofthescientificcommu- scientists, and particularly systematic reviewers, closer
nity to utilize theresourcesavailable most efficientlywhen internalization of research and collaboration. Similarly, the
exploring research priorities to afford the health needs of creation of some international collaborations, including
the population, stimulating north–south and west–east those conducting clinical trials, may have settled the
collaborations where possible. In fact, these results are groundwork for the subsequent realization of collaborative
consistentwiththosereportedbyUthmanetal.[26],who meta-analyses that may have a clear scientific and clinical
assessed the characteristics of the 100 most frequently impact. For example, according to SCI-E, the most cited
cited meta-analysis related articles. Although the scope of meta-analysisarticlehasreceivedmorethan2,500citations;