Randomized clinical trials in orthodontics are rarely registered a priori and often published late or not at all

RESEARCH ARTICLE Randomized clinical ials in o hodon ics a e a ely egis e ed a p io i and o en published la e o no a all Spy idon N. Papageo giou 1 *, Geo gios N. An onoglou 2,3 , Geo ge K. Sa ´ndo 2,4 , Theodo e Eliades 1 1Clinic o O hodon ics and Pedia ic Den is y, Cen e o Den al Medicine, Uni e si y o Zu ich, Zu ich, Swi ze land, 2Ins i u e o Den is y, Depa men o O al and Maxillo acial Su ge y, Uni e si y o Oulu, Oulu, Finland, 3Depa men o Pe iodon ology and Implan Biology, Den al School, A is o le Uni e si y o Thessaloniki, Thessaloniki, G eece, 4BioMediTech, Ins i u e o Bioscience and Technology, Uni e si y o Tampe e, Tampe e, Finland *[email p o ec ed] Abs ac A p io i egis a ion o andomized clinical ials is c ucial o he anspa ency and c edibili y o hei indings. Aim o his s udy was o assess he equency wi h which egis e ed and comple ed andomized ials in o hodon ics a e published. We sea ched ClinicalT ials.go and ISRCTN o egis e ed andomized clinical ials in o hodon ics ha had been com- ple ed up o Janua y 2017 and judged he publica ion s a us and da e o egis e ed ials using a sys ema ic p o ocol. S a is ical analysis included desc ip i e s a is ics, chi-squa e o Fishe exac es s, and Kaplan-Meie su i al es ima es. F om he 266 o hodon ic ials eg- is e ed up o Janua y 2017, 80 ials had been comple ed and included in he p esen s udy. Among hese 80 included ials, he majo i y (76%) we e egis e ed e ospec i ely, while only 33 (41%) we e published a he ime. The median ime om comple ion o publica ion was 20.1 mon hs (in e qua ile ange: 9.1 o 31.6 mon hs), while su i al analysis indica ed ha less han 10% o he ials we e published a e 5 yea s om hei comple ion. Finally, 22 (28%) o comple ed ials emain unpublished e en a e 5 yea s om hei comple ion. Publica ion a es o egis e ed andomized ials in o hodon ics emained low, e en 5 yea s a e hei comple ion da e. In oduc ion Randomized clinical ials a e he gold s anda d in compa a i e e ec i eness esea ch, due o hei explici me hods, in e nal alidi y, and anspa ency. C ucial o his anspa ency is he p o ision o an a p io i designed p o ocol ha delinea es each ial aspec . This can also be used o compa e he o iginal ial plan wi h i s subsequen publica ion [1] o asce ain he i- al’s o iginal scope, as well as minimize he isk o da a d edging. Fu he mo e, a p io i egis a- ion o clinical ials can sa egua d agains dange s like delayed publica ion o non-publica ion o ials, selec i e epo ing o ou comes, pe p o ocol a he han in en ion- o- ea analyses, PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0182785 Augus 4, 2017 1 / 13 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Ci a ion: Papageo giou SN, An onoglou GN, Sa ´ndo GK, Eliades T (2017) Randomized clinical ials in o hodon ics a e a ely egis e ed a p io i and o en published la e o no a all. PLoS ONE 12 (8): e0182785. h ps://doi.o g/10.1371/jou nal. pone.0182785 Edi o : Pe e M. Milg om, Uni e si y o Washing on, UNITED STATES Recei ed: May 16, 2017 Accep ed: July 24, 2017 Published: Augus 4, 2017 Copy igh : ©2017 Papageo giou e al. This is an open access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal au ho and sou ce a e c edi ed. Da a A ailabili y S a emen : All ele an da a a e wi hin he pape and i s Suppo ing In o ma ion iles. Funding: No unding o disclose o his s udy. Compe ing in e es s: The au ho s ha e decla ed ha no compe ing in e es s exis . and ial o e laps among pape s included wi hin sys ema ic e iews [1–4]. Such a p io i egis- e ed ial p o ocols can be ei he published as s andalone publica ions [5] and/o be egis e ed in eely accessible online eposi o ies [6]. One such eely accessible online eposi o y is ClinicalT ials.go (h ps://clinical ials.go /), a ial egis y and esul s da abase main ained by he US Na ional Lib a y o Medicine. Al hough he e is no o mal equi emen ha all unded ials mus be egis e ed in Clinical- T ials.go , he In e na ional Commi ee o Medical Jou nal Edi o s (ICMJE) began om 2005 equi ing ial egis a ion as a p e equisi e o publica ion in any o i s membe jou nals [7]. This led o a conside able inc ease o ials being egis e ed, which now each se e al hund eds o new ials being egis e ed each week [8]. Ano he well-known egis y o clinical ials in medicine is he In e na ional S anda d Randomised Con olled T ial Numbe (ISRCTN) eg- is y (h p://www.is c n.com/), which con ains he basic se o da a i ems deemed essen ial o desc ibe a s udy a incep ion, as pe he equi emen s se ou by he Wo ld Heal h O ganiza- ion, he In e na ional Clinical T ials Regis y Pla o m, and he ICMJE guidelines. These wo egis ies ep esen he wo majo ecipien s o ial egis a ions wo ldwide [8], while all ial eco ds in hese da abases a e eely accessible and ha e been assigned a unique ial iden i ie , he eby enabling linking ial egis a ions o hei subsequen publica ions. Howe e , exis ing esea ch shows ha he esul s o s udies a e o en no sha ed publicly in a imely way and ha be ween 25% and 50% o clinical ials emain unpublished e en se e al yea s a e comple ion [9–11]. The e a e many possible easons behind he delayed o non- publica ion o esul s om clinical ials, including lack o incen i e o dissemina e nega i e indings, ime cons ain s, limi ed esou ces, changing esea ch in e es s, o e en ailu e o ha e an a icle accep ed by a jou nal. Impo an ly, ime-lags in he dissemina ion o esul s may ha e ad e se consequences o he p ac ice o e idence based medicine and o public heal h [12]. The non-publica ion o ial esul s also iola es an e hical obliga ion ha ial in es iga o s ha e owa ds ial pa icipan s [13]. E en i ials do e en ually ge published yea s la e , i may be oo la e, and he esul s may ha e less ele ance because o he apid changes in he landscape o e idence. Al hough se e al s udies ha e es ima ed he p opo ion o incomple e o selec i e epo ing in a ious medical special ies [14–17], no such assessmen s exis in o hodon ics. A p e ious s udy [18] p o ided an o e iew o egis e ed ials in o hodon ics up o 2014, bu was limi ed o desc ip i e analysis o hei egis a ions and did no assess which we e e en ually pub- lished. Aim o he p esen s udy was o assess he publica ion a e o egis e ed o hodon ic andomized clinical ials and o assess cha ac e is ics ela ed o hei publica ion iming. Ma e ials and me hods S udy sample The p o ocol o he p esen s udy was d a ed p io o s udy ini ia ion, bu was no egis e ed in PROSPERO, as i did no mee PROSPERO’s p e equisi e o ha ing clinical connec ion o he ials’ esul s. Fo his s udy wo au ho s (SNP and GNA) sea ched independen ly wo online egis ies o andomized ials, ClinicalT ials.go and ISRCTN, up o Janua y 2017. As we in ended o assess he publica ion a e o hese ials, only egis a ions o ials ma ked in he egis ies as comple ed we e included. Regis a ions o non- andomized o non-o hodon- ic ials we e excluded, while any egis a ions pe aining o he same ial we e g ouped oge he . The same wo au ho s (SNP and GNA) subsequen ly sea ched independen ly in Janua y 2017 o any publica ions o igina ing om he comple ed egis e ed ials using a p e ious comp ehensi e loca ing p o ocol [19]. Ini ially, he publica ion ield o he ial p o ocol Regis a ion and publica ion o ials in o hodon ics PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0182785 Augus 4, 2017 2 / 13 egis a ion was sea ched o any linked e e ences. Secondly, he unique ial iden i ie assigned wi hin each egis a ion was used o sea ch he web h ough Google (h ps://www. google.com), Google Schola (h ps://schola .google.com/), and MEDLINE ia Pubmed (h ps://www.ncbi.nlm.nih.go /pubmed/). Finally, i no co esponding publica ions could be aced, he names o all in es iga o s lis ed in he ial p o ocol as p incipal in es iga o o o h- e wise we e used o sea ches in MEDLINE, combining hem wi h he keywo d o hodon (Table A in S1 File). Manually-iden i ied publica ions we e judged as o igina ing om a spe- ci ic ial egis a ion, i ei he (i) he unique egis a ion iden i ie was epo ed wi hin he ial publica ion o (ii) he used au ho names o e lapped be ween p o ocol and publica ion, he ial was andomized and on he same subjec as he p o ocol, and pa ien en ollmen in he publica ion ma ched he ial p o ocol desc ip ion. Only publica ions om scien i ic jou - nals we e included, while ial publica ions in he o m o heses/disse a ions we e excluded, as hey ollow conside ably di e en publica ion pa hways and a e no subjec o he same pee e iew p ocedu es as jou nal pape s. The ial’s Pubmed Unique Iden i ie (PMID) (o manual unique iden i ie s in case ha didn’ exis ) was assigned o each iden i ied ial publica ion o iden i ica ion pu poses. Da a ex ac ion The same wo au ho s (SNP and GNA) independen ly ex ac ed da a on he le el o ial egis- a ion and ial publica ion, using p e-de ined and calib a ed ex ac ion o ms. Ex ac ed da a om he ial’s egis a ion included egis y, egis a ion numbe , clinal se ing (uni e si y clinic, hospi al, o p i a e p ac ice), geog aphic in o ma ion (coun y and con inen ), ial s a - and end-da e, sponso (in e nal, go e nmen , comme cial, o o he ), egis a ion’s im- ing ( egis e ed be o e o a e ial ini ia ion [20]), ial na u e (single-cen e o mul icen e ), ou come ype (pa ien - epo ed o in es iga o -assessed), age o eligible pa ien s (child en, adul s, o mixed), and ial size (a bi a ily ca ego ized as o <100 pa ien s pe ial a m based on empi ical da a [21]). Ex ac ed da a om he ial’s publica ion included, PMID, pub- lica ion da e, numbe o iden i ied publica ions om he same ial, jou nal name, jou nal publica ion ype (wi h o wi hou elec onic publica ion ahead o p in ), and ial indings (judged om he abs ac as posi i e, no posi i e, o unclea ). In o ma ion abou he a ilia- ion and geog aphical loca ion o a ial was based on he pe son lis ed as p incipal in es iga o o each ial o he esponsible pa y, i no p incipal in es iga o was lis ed. We measu ed he ime om ial comple ion o publica ion by calcula ing he du a ion o ime (in mon hs). The ial comple ion da e was de e mined om he ial’s egis a ion, while he ea lies da e a ial eme ged in MEDLINE was adop ed as i s publica ion da e. Fo wo ials ha we e no lis ed in MEDLINE publica ion da e was he i s day o he jou nal publica ion issue’s mon h. I a ial publica ion pe ained o p elimina y esul s o a ial and no he inal esul s o he ial’s p ima y ou come, he ial was coded as unpublished. Fo all ials wi hou any iden i ied pub- lica ions we calcula ed he ollow-up ime as he du a ion in mon hs be ween ial comple ion and end da e o ou s udy (Janua y 12, 2017). S a is ical analysis We conduc ed a desc ip i e analysis calcula ing equencies o bina y ou comes o medians, In e qua ile Ranges (IQRs), and anges o con inuous ou comes, due o skewness o he la e . Simple di e ences in he ial publica ion a e and comple ion- o-publica ion ime acco ding o he a ious desc ip i e cha ac e is ics we e in es iga ed wi h chi-squa e /Fishe exac and K uskal-Wallis es s, espec i ely. Fo he analyses, geog aphic egions (con inen ) we e me ged oge he , i less han 10 ials o a oid da a hinning. In addi ion, we e alua ed Regis a ion and publica ion o ials in o hodon ics PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0182785 Augus 4, 2017 3 / 13 publica ion a e o included egis e ed ials using Kaplan-Meie analysis a e checking o assump ions and assessed in luencing ac o s wi h g aphical inspec ion and log- ank es s. Fo he su i al analyses he comple ion- o-publica ion ime o published ials (o he com- ple ion- o-Janua y 2017 ime o unpublished ials) we e used as ollow-up pe iods, while all ials ha emained unpublished we e censo ed. All analyses o da a (S1 Da ase ) we e pe - o med in S a a e sion SE 14 (S a aCo p LP, College S a ion, TX) wi h a le el o signi icance se a 5%. Resul s We iden i ied a o al o 266 egis e ed ials in o hodon ic, 130 o which we e comple ed in Janua y 2017, and we e assessed o eligibili y in he p esen s udy (Fig 1; Table B in S1 File). Finally, a o al o 50 egis e ed ials we e judged as ineligible, lea ing 80 egis e ed ials ha we e inally included in his s udy. These pe ained o 80 andomized ials, he majo i y (59%) o which o igina ed om Eu ope and speci ically om he Uni ed Kingdom (44%). They we e ini ia ed be ween 1997–2015 and we e comple ed be ween 2003–2016 (Table 1). F om he iden i ied ials, hal o hem (50%) had no ex e nal inancial suppo , while only he mino i y (8%) included mul iple ial si es (Table 2). The median numbe o planned pa ien s o be en olled in he ial was 50 pa ien s (IQR: 30–90 pa ien s; Table 3), while only 23% o he ials had a la ge sample size (100 pa ien s). In e es ingly, only 24% o ial p o o- cols we e egis e ed p ospec i ely and he as majo i y o ials (76%) we e egis e ed a e ial ini ia ion. As a as publica ion a e o he egis e ed ials is conce ned, only 33 (41%) o he 80 iden i- ied comple ed ials including a leas 3839 pa ien s had been published in jou nals a he ime his s udy was comple ed, wi h 10% o ials being published in mul iple pape s (Table 2). The majo i y o egis e ed ials we e published in special y jou nals and in jou nals ha p o ided elec onic publica ion ahead o p in op ion (64% and 55% espec i ely). Finally, abou equal pa s o he 33 published ials epo ed posi i e and non-posi i e indings (42% and 45%, espec i ely), while esul s we e inconclusi e in he emaining 4 ials (12%). No associa ion be ween publica ion a e and cha ac e is ics o he egis e ed ials could be ound (Table 3). The single excep ion was he geog aphic loca ion o he ial, wi h egis e ed ials om No h Ame ica and A ica being published less o en han ials om o he con inen s. Howe e , he emaining 47 (59%) egis e ed and comple ed ials, including a leas 2355 andomized pa ien s, emained unpublished a he ime o he p esen s udy. Among he 33 comple ed ials ha we e e en ually published, 10 (30%), 18 (55%), 27 (82%), 29 (88%), and 30 (91%) o hem we e published wi hin 12, 24, 36, 48, and 60 mon hs a e ial comple ion, espec i ely (Table 2). The median ime om comple ion o publica ion was 20.1 mon hs wi h an IQR o 9.1–31.6 mon hs) (Table 4). Conside able di e ences in he comple ion- o-publica ion ime we e ound acco ding o egis y, sponso ype, pa ien age, and jou nal (Table C in S1 File), wi h ials om ClinicalT ials.go , in e nally sponso ed, wi h adul o mixed pa ien popula ions, and ials submi ed in non-special y jou nals we e pub- lished as e han he es . The Kaplan-Meie es ima es o he publica ion a e o he 80 included ials indica ed ha mos o hem we e published du ing he i s 24 o 48 mon hs a e comple ion (Figu e A in S2 File). On he o he side, abou 60 mon hs ( i e yea s) a e comple ion no conside able publi- ca ion ac i i y was seen. Di e ences in he su i al cu es can be seen in Figu es B-M in S2 File. T ials egis e ed in ClinicalT ials.go seemed o be published as e and mo e o en han ials egis e ed in ISRCTN (Figu e B in S2 File), while conside able di e ences in he publica- ion a e we e seen ac oss con inen s (Figu e E in S2 File). Addi ionally, ials wi h adul Regis a ion and publica ion o ials in o hodon ics PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0182785 Augus 4, 2017 4 / 13 pa ien s o ials wi h unspeci ied pa ien popula ions seemed o be published less o mo e slowly han ials on child en (Figu e I in S2 File). Finally, among he 33 published ials, ials in non-special y jou nals seemed o ge published as e han ial publica ions in special y jou nals (Figu e K in S2 File). As he majo i y (30 ou o 33 ials; 91%) o comple ed and published ials we e published wi hin 5 yea s om hei comple ion (Table 2), a ee-pe iod o 5 yea s om comple ion was allowed o all 80 iden i ied and comple ed ials. Howe e , abou one hi d (n = 22; 28%) o comple ed ials s ill emained unpublished, e en hough 5 yea s had elapsed om ial com- ple ion, wi h small chances o he ial inally ge ing published. Fig 1. Flowdiag am o he s udy selec ion. h ps://doi.o g/10.1371/jou nal.pone.0182785.g001 Regis a ion and publica ion o ials in o hodon ics PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0182785 Augus 4, 2017 5 / 13 Discussion Summa y o e idence The p esen s udy assessed he publica ion a e as well as he comple ion- o-publica ion ime o all andomized ials in o hodon ics ha had been egis e ed in ClinicalT ials.go and ISRCTN and had been iled as ‘comple ed’ up o Janua y 2017. The esul s indica ed ha a he ime o comple ion o he p esen s udy (Janua y 2017) 59% o all egis e ed and comple ed ials had no been published, which co esponded o a leas 2355 included pa ien s (some ials did no epo he planned pa ien sample). This ag ees wi h simila s udies on he biomedical li e a u e ha epo non-publica ion a es o 29% [15], 34% [16], 50% [17], and 54% [14]. As a as ime o publica ion is conce ned, he esul s indica ed ha he median comple- ion- o-publica ion ime was 20.1 mon hs (1.6 yea s). This also is simila o he indings o p e- ious in es iga ions, which epo a median comple ion- o-publica ion ime o 28.8 mon hs (2.4 yea s) [22]. Addi ionally, su i al analysis indica ed ha mos o he ials we e published Table 1. Demog aphics o he included egis e ed ial p o ocols. N a N % Regis e ClinicalT ials.go 80 45 56% ISRCTN 35 44% Coun y Aus ia 80 1 1% Belgium 1 1% B azil 5 6% Canada 2 3% China 1 1% Colombia 1 1% Egyp 1 1% F ance 2 3% Ge many 4 5% Hong Kong 1 1% India 3 4% I an 1 1% I aly 2 3% Kuwai 1 1% Sweden 1 1% Sy ia 4 5% Uni ed Kingdom 35 44% Uin ed S a es o Ame ica 12 15% No epo ed 2 3% Se ing Uni e si y 80 63 79% P i a e 1 1% Hospi al 16 20% Con inen Eu ope 80 47 59% Asia 11 14% No h Ame ica 14 18% Sou h Ame ica & A ica 8 10% ISRCTN = In e na ional S anda d Randomised Con olled T ial Numbe ; N a = eligible ials ials included in he assessmen o his ac o ; N = ials in each ac o ca ego y. h ps://doi.o g/10.1371/jou nal.pone.0182785. 001 Regis a ion and publica ion o ials in o hodon ics PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0182785 Augus 4, 2017 6 / 13 Table 2. Cha ac e is ics o he included egis e ed ial p o ocols (bina y a iables). Fac o Ca ego y N a N % Sponso In e nal 80 40 50% Go e nmen 21 26% Comme cial 10 13% O he 9 11% Regis a ion P ospec i e 80 19 24% Re ospec i e 61 76% Ou come Pa ien - epo ed 79 6 8% In es iga o -asse ed 73 92% La ge ial (100 pa ien s pe ial a m) Yes 71 5 7% No 66 93% Mul icen e ial Yes 80 6 8% No 74 93% Eligible pa ien s Child en 80 29 36% Adul s 8 10% Mixed 29 36% No speci ied 14 18% T ial published Yes 80 33 41% No 47 59% Publica ion a e a e ial comple ion Wi hin 12 mon hs 33 10 30% Wi hin 24 mon hs 18 55% Wi hin 36 mon hs 27 82% Wi hin 48 mon hs 29 88% Wi hin 60 mon hs 30 91% Publica ions om egis e ed ial p o ocol None 80 47 59% One 25 31% Two 5 6% Th ee 3 4% T ials emaining unpublished a e 5 yea s Yes 80 22 28% No 58 72% Jou nal Am J O hod Den o acial O hop 33 8 24% Angle O hod 1 3% B J O al Maxillo ac Su g 1 3% Ca ies Res 1 3% Eu J O hod 6 18% In J Paedia Den 1 3% J Am Den Assoc 1 3% J Clin Diagn Res 1 3% J Con emp Den P ac 1 3% J Den Res 3 9% J Indian O hod Soc 1 3% J O hod 5 15% J Pha m Biomed Sci 1 3% Pho omed Lase Su g 1 3% Sci Rep 1 3% Jou nal ype Special y (o hodon ic) 33 21 64% Non-special y 12 36% Jou nal wi h E-publica ion ahead o p in Yes 33 18 55% (Con inued) Regis a ion and publica ion o ials in o hodon ics PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0182785 Augus 4, 2017 7 / 13 wi hin he i s 24–48 mon hs (2–4 yea s) a e he ial’s comple ion. A e abou 60 mon hs (5 yea s) he ial publica ion a e was minimal, wi h only 3 ials (9% o all published ials) being published his ime poin . This is simila o he esul s o Chapman e al. [16] ha indi- ca ed ha he majo i y o published ials (73%) ge s published wi hin 48 mon hs a e ial comple ion. This seems o also ag ee wi h Li as e al. [23] epo ing ha only hal o s udy abs ac s p esen ed in o hodon ic cong esses ( he e o e pe aining o comple ed s udies) Table 2. (Con inued) Fac o Ca ego y N a N % No 15 45% T ial esul s Posi i e 33 15 42% No posi i e 14 45% Unclea 4 12% N a = eligible ials ials included in he assessmen o his ac o ; N = ials in each ac o ca ego y. h ps://doi.o g/10.1371/jou nal.pone.0182785. 002 Table 3. C oss- abula ion o subsequen ial publica ion and p o ocol cha ac e is ics. Fac o Ca ego y Published N (%) Unpublished N (%) P* Regis y ClinicalT ials.go 20 (44%) 25 (56%) 0.51 ISRCTN 13 (37%) 22 (63%) Sponso In e nal 14 (35%) 26 (65%) 0.30 Go e nmen 10 (48%) 11 (52%) Comme cial 3 (30%) 7 (70%) O he 6 (67%) 3 (33%) A ilia ion Uni e si y 26 (41%) 37 (59%) 1.00 P i a e p ac ice 0 (0%) 1 (100%) Hospi al 7 (78%) 9 (22%) Con inen Eu ope 20 (43%) 27 (57%) 0.30 Asia 8 (73%) 3 (27%) N. Ame ica 2 (14%) 12 (86%) S. Ame ica & A ica 3 (38%) 5 (62%) Regis a ion P ospec i ely 26 (43%) 35 (57%) 0.66 Re ospec i ely 7 (37%) 12 (63%) Ou come Pa ien - epo ed 31 (42%) 42 (58%) 1.00 Doc o -assessed 2 (33%) 4 (67%) Mul icen e ial Yes 2 (33%) 4 (67%) 1.00 No 31 (42%) 42 (58%) Pa ien s Child en 14 (48%) 15 (52%) 0.07 Adul s 2 (25%) 6 (75%) Mixed 15 (52%) 14 (48%) No speci ied 2 (14%) 12 (86%) La ge ial Yes 4 (80%) 1 (20%) 0.17 No 28 (42%) 38 (58%) ISRCTN = In e na ional S anda d Randomised Con olled T ial Numbe . * om chi-squa e (o Fishe ’s exac es , when mo e han 20% o he cells ha e expec ed equency less han i e o a leas one is null) h ps://doi.o g/10.1371/jou nal.pone.0182785. 003 Regis a ion and publica ion o ials in o hodon ics PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0182785 Augus 4, 2017 8 / 13 each ull publica ion, wi h a median p esen a ion- o-publica ion ime o 16 mon hs (1.3 yea s). Se e al egis y-, ial-, o pa ien - ela ed cha ac e is ics seemed o a ec he publica ion a e o included egis e ed ials, as seen h ough he Kaplan-Meie su i al es ima es and he log- ank es s (Table D in S1 File; Figu es B-M in S2 File). Fo one, conside able di e ences we e obse ed be ween ials egis e ed in ClinicalT ials.go and hose egis e ed in ISRCTN, wi h ials egis e ed in he la e being published less o en and la e . This migh indica e ha a ue di e ence exis s in he publica ion o comple ed ials, bu i migh also be associa ed wi h he conside able geog aphic di e ences in he con en o he wo egis ies, as ISRCTN included ials p ima ily om he Uni ed Kingdom, while ClinicalT ials.go had a wide con- ibu ion spec um (Figu e N in S2 File). Coun ing only unpublished ials wi h a ollow-up pe iod o mo e han 5 yea s a e ial comple ion o allow o a ee pe iod o be published, indica ed ha 28% (n = 22) o iden i ied ials s ill emained unpublished. This is o impo ance as, based on p esen esul s, he chance o hose ials being e en ually published is small. This ep esen s a was e o esou ces as well as a gap in he exis ing esea ch e idence in o hodon ics ha can ha e a di ec impac on sys- ema ic e iews o clinical ials and subsequen ly on clinical decision making. Fo example, se e al ials on new appliances o no el adjunc p ocedu es aimed o expedi e o hodon ic ea men emain o he p esen unpublished (Table E in S1 File), which means ha exis ing sys ema ic e iews on his subjec [24–27] migh no e lec he ac ual e idence base. The e a e se e al s eps ha could be implemen ed owa d imp o emen o ial egis a ion and ial publica ion p ac ices. Agencies unding clinical ials could implemen s ic e su - eillance sys ems o ial publica ion a e and u ge esea che s o publish he esul s o hei ials, no ma e how posi i e o nega i e hese migh seem. Au ho s a e encou aged o egis e ials p io o ial ini ia ion. In he p esen s udy, he majo i y (76%) o andomized ials we e egis e ed e ospec i ely, some hing ha has also been seen in o he ields o medicine [20], bu does nei he co espond o he no ion o a p io i ial egis a ion no does i sa e- gua d agains selec i e epo ing. Edi o s o o hodon ic jou nals should a oid judging ial submissions om he a ac i eness o hei indings and should pu sue publica ion o ials wi h nega i e esul s. Fu he mo e, edi o s can indi ec ly p omo e ial egis a ion by using i as a p e equisi e o a ial o be conside ed o publica ion in he jou nals hey manage. The ICMJE changed policy in 2005 o adop his [7], which esul ed in an inc ease o ial egis a- ion [8]. This equi emen was la e inco po a ed in o he ICMJE’s “uni o m equi emen s o manusc ip s submi ed o biomedical jou nals,” along wi h he upda ed CONSORT 2010 s a e- men o he epo ing o andomized con olled ials [28]. E en hough a la ge pa o medi- cal jou nals equi es ials o be egis e ed [29], he landscape in o hodon ics s ill emains bleak: al hough hal (5 ou o 10) o o hodon ic jou nals men ion egis a ion o andomized clinical ials a hei ‘Ins uc ions o au ho s’ websi es, only one (Jou nal o O hodon ics) wo ds his explici ly as a equi emen o publica ion. Table 4. Cha ac e is ics o he included egis e ed ial p o ocols (con inuous a iables). Ou come Ex ac ed om N a Mean (SD) Median (IQR) Range Repo ed sample size (pa ien s) P o ocol 71 87.2 (138.4) 50.0 (30.0–90.0) 8.0–1000.0 T ial du a ion (mon hs) P o ocol 80 29.5 (22.4) 24.3 (15.2–36.5) 0.9–124.8 Comple ion- o-publica ion (mon hs) P o ocol / publica ion 33 24.5 (18.0) 20.1 (9.1–31.6) 1.8–65.0 Mean ollow-up o ial p o ocol o he p esen s udy (mon hs) P o ocol / publica ion 80 47.7 (45.8) 28.4 (13.0–64.7) 1.8–166.6 SD = s anda d de ia ion; IQR = in e qua ile ange (Q1-Q3). h ps://doi.o g/10.1371/jou nal.pone.0182785. 004 Regis a ion and publica ion o ials in o hodon ics PLOS ONE | h ps://doi.o g/10.1371/jou nal.pone.0182785 Augus 4, 2017 9 / 13