Tobacco Documents Online

rrom' acan sk rax A o- Loreen wicaipin NOTICE TFpS MATR/AL ldAY SE PKOTECTED BY CrJPYRtGHT LAW `TT? E I 7 U.S. C~Gf) tas- dnk tin d 37e- ~oao. .~ dec- i4•7. sd J 1 ~ : Source of Funding and Outcome of Clinical Trials RICHARD A. DAVIDSON, MD, MPH' D.cau.e o( r.oent ccac.rar about coalifah of latats.t aad prbUai.d r+s.areb. the artbar a.aysed JQ7 ceaholl.d cUntca/ trfaL. 3tudf.s were claads.d as (a.oriap.ftia a s.+w tb.rarFr or a tmdltioaal tb.wpy. and as b.lsp sr~ pxt.j by a pbarmoeaatktitl =z*ufacftu%w er aa belag Ora- .ec11T sappotted. S.v..lt•ms p.r e.st of tbe tr9da lv.at.d sew tb.ev*.:: 43% e! tb....a0 tasd.d b7 piazaac~..- tical ti..ow OI the J1 Mals fa.ortap hodiH.ed tL.evpt. .ay foat (19%t wae iupporl.d b7 a pharaenslfcal tErom. 7iee* was a stoltstteQly .fp.ificant aarociattoa b.ews.a the aotrce d(oadisp and the oatcome el the stStdy (p a 8.0QL. Few trlala ssppoat.d by pharaaac.utlcal .m.afoc- tsarrs favorrd leedttlasal tfna"r soese no.aas tas fhts ![adfap n4t 1aclads s.f.attm at drugs HEsy to 6e psoms .(11eoe4.aa 7Tps ll.eaa (lzta.-..*atlve atadN.J. and t.ar .t d~oattaeatiea of fuadtae sbeuld atcli atadl.s be sub- aeftt.d. Imporfaet ~tlnleal ialorsathn mar be l.rt !t s.r- a?1v, rt.dlr aam e.t pubJlsirrd. ty .roe&s: etlnieal trfal.: (und1sgr pLaQSOe.attcal ma,autecturer support. I Get Lt. mut 1Ra>r ilM:1:155--1S8. CoUtRO11.m ct.uttCAt. TRtuA provide the strongest svidence that a therapy is efficacious. The number o( these trials has increased in the medical liter- atun.I With recent limitations of government fund- iag. pharmaceutical firms increasingly have .ponsored this relatively costly research. For ex- ample, one study of members of the Infectious Dis- .as. Society= found that in 1977 more investiQators from the Society reportedd receiving funds from in- dustry than from the National Institutes of Health {39X vs.1696, respectively); over half of the respond- ents thought that support from pharmaceutical firms was necessary. Research funding by the pharmaceutical in- dnstry raises the possibility of couflicts of interest. Suggestions for dealing with this problem include disttibution of Industry fuads through independent sponsors, restrictions of paid consultations, limi- tation of entertainment and gifts, abandonment of 1be funding of categorical speakers for medical aehools, and reatrictions on stock holdings by in- *estigatora? The publication of symposia has also ,aised concern about sponsorship and the objec- rivity of research because they may provide infor- mation that has not been peer-reviewed and may iail to invite investigators whose results are re- 9arded as unfavorablo.3-' Studies that undergo critical review in the ttandard publication process are generally consid- a'ad to be free of special interest bias. In spite of 'r'WIO~ PraStmr. Oeputrnent of Medicine. Box J2T7. Untversity of fiorida. Cainesville. Fiorida 32610. Supporfsd !n part by Grant NIH #PE 14233. Addr2ss corrEspondence and reprint requests to Dr. Davitlson, this view, the New England Journal of Medicine chose to strenflth.n its requirements regarding the disclosure of possible conflicts of interest between investiqators and industry.', ' $ecaua.little is known about whether and how industry sponsorship affects published rss.arch. I undertook an analysis of the results of clinical trials according to funding source. METHODS I reviewed all clinical trials that included a concurrent or cross-over control group and were published in 1964 in the following journals: the New England Journal of Mediciae, Annals of Internal Medicine, the American Journal of Medicine, 11z- chives of Internal Medicine, and the Lancet. These were selected becaus. they represent the joarnals from which int.rnal medicine practitioners obtain much therapeutic infacmation. Many diffm+.nt kinds of thsapies rrow inc2vd.d. Studies that pooled data from other trials, measured no clinical outcomes, or were published In a symposium were excluded. Two bsfore-oad-after studies were included be- cav.se of a short interval between baseline and post- intervention evaluation. Studies wa:+e dassifisd as either favorinq a new therapy or intervention or favoring traditional man- agement. This classification was chosen over a simple dichotomy into positive and n.Qative stud- ies because in some instances a new therapy that is, for example, less expensive might actually be favored by a study that found no difference in th.r- apies (a "neQative" study). The cabegotisation was done without knowledge of the source of fundin{7. The classification of results was first attempted by reviewing the abstract alone. This was occanion- ally difficult because some studies did not clearly state their hypotheses with regard to specific out- come measures; this sometimes led to a situation in which, for instance, mortality might be un- chtmQed, but various markers of morbidity were redueed.1° Occasionally a study ssemed to fa.or both traditional and now therapies.11 In these few situations, the entire article was carefully re- viewed and ths study .vas classified according to the autho:'s recommendations. Information conceraing the trials was collected from the "Methods" section. The sarrlple size pes group was calculatedd for groups of unequal size by dividing th. total number of patients in the trial by the number of groups; the presence or absence -of blinding was recorded-

rromacan s rax i o- Loreen ivicaipin 4 1ss 7S(71 %) Favored new therapy 33(43%) Phanrxacsutiaaliy supported un~c. . ...,... ~ .. . ....... ....ti....... j,aS/%Ql0I7. SOUPCE OF FUNIXNG ANO SiUOY U7TCOME 107 trials 31(25%) Favorod traditional therapy 43(57%) 403%) Q.w.ra/!y fh.rssa..rt)cally supported suOPortad Chl sq.r. = 9.045, p =.002 Trials were then classified by funding source. A trial was considered "pharmacwutically sup- pocted,, if a drug company .vas noted in the ac- kisowled9ements as having provided support for the study or any of its investigators (a study was not classified as pharmaceutically supported if a drug company provided only drugs or placebos). All other studies were classified as generally sup- ported, whether funding sources were noted or not. Statistical analysis was by chi square. RESULTS One hundred and twenty-six trials were re- viewed; 107 trials met the final inclusion criteria. Of the remaining 19 triaL, 13 were published in symposia, two were actually caselcontrol studies, two used pooled data, one measured no clinical outcomes, and one compared different dosages of the same drug. There was a statistically significant associa- tion between the source of funding and the outcome of the study (p = 0.002), with studies supported by pharmaceutical manufacturers favoring new ther- apies in comparison with generally supported studies (Fig. I). The distribution of studies by jour- nal, results, and funding source is shown in Table 1. The ratios of articles favoring new to those fa- voring traditional therapies varied from 6 to 1 in the Annals of Internal Medicine to 1 to 1.4 in the American Journal of Medicine. Twenty-two studies evaluated non-pharma- cologic therapies such as sclerotherapy," hypno- therapy, t' or oxorcise." If such studies more fre- quently had negative results, the analysis could have been biased, since phoanaceutical firms are much more likely to fund studies of pharmaceutical agents. In fact, 60% of studies of non-pharmaco- logic therapies favored the new therapy in com- parison with 71% of the overall group, and only one 2 7(a7%) Gen.ra)ly srpported 17pucs t. D=atptla of 107 aInlal trtslsk of the 22 was pharmaceutically supported. For this reason. an analysis that included pharrnacologic studies only was prepared. Tha strength of the as- sociation between funding and results dscreassd somewhat but remained statistically significant (p = 0.007). Analysis of sample size per group revealed no significant differences between the pbazmac.ufi- cally supported and generally supported studies (Table 2). With regard to blinding, 67.5% of the pharmaceutically supported studies specificallY mentioned that blinding was carried out, compared with only 41.8% of the generally supported studies (p - 0.01). However, this was a reflection of the non-pharmacologic studies, which utilized inter- ventions that could not easily be blinded. Analysis (PhG outc f.ct alon souz ship !ry f clast Imn. 3tud: that conh what ical t tria2. ment less studi thesw ative smai error snt).' tors. samF to de feren journ icant mace P sults of pharmacologic studies alone revealed no sig- I Sour nificant difference in rates of blinding between the two funding sources. (p = 0.46). Among the four pharmaceutically supported studies favoring traditional therapies, one sug- gosted that the therapy, v.hile not improving over- all survival, ". .. may be helpful early in the course of the disease and in c.rtaia subgroups of pa- tients:'10 Another compared two therapies mcmu- factured by the sponsoring drug company.'S The other two studiee, one comparing the lipid-reduc- ing effects of d- and 1-thyroxine," and the other a decision-analysis study evaluating the cost-eff.o- tiveness of two antibiotics," were funded by com- panies that did not produce any of the drugs investigated. In no case wai a therapeutic agent manufactured by the sponsoring company found to *w Eij PS G5 txXar PS GS Mnals a PS QS PS c= be inferior to an alternative product manufacturQd ar by another company. ~ ~ .~ DISCUSSION I - r.) Ln ~~ This study has demonstrated a statisticaitf ~ ~ significant association between source of fundinQ ncrnt„ na

~ ~ d- JauRW oF GExcs+u tecrMAL t.teflKrrt. Vdume 1(AIayUune). 15196 (phcamaceutical firm versus Qeneral support) a.nd outcome of published clinical trials. Cause and ef- f.et cannot be inferred on the basis of this study aion:. Other factors associaied with both funding source and outcome may confound the relation- ship. Also, it is possible that the chatacterisation by funding source may have inadvertently mis- datsified some studi.s becaus. of inaccurate ac- knowledgements, underestimating the number of studies funded by pharmaceutical firms. "Publicati.on biaa." in this case a gceata chance that a study will be accepted for publication if it contains a positive result, may d.termine, in part, what appears in journals. One discussion of clin- ical triaL'' suggested that for every clearly positive triai, there are five to ten trials comparing treat- ments that o=e equally effective. It is possible that 1as well-known journals might publish negative studies more frequently, and thus ha.* more of th.se studies submitted. Furthermore, many neg- ative studies are methodologically weakened by small sample sises and the possibility of a Type II error (the inability to detect a true difference if pres- •nt).1r This may make them less attractive to edi- toss. It is possible that funding may influence sample sise, which in turn might aff.ct the ability eo detect clinically or statistically significant dif- ieseaces, or the chance of publication in a major journal. In fact, in this study there were no siQnif- icant differences in sample siae between phar- maceutically and generally supported studies. Publication bias alone cannot explain the re- sults of the current study, as there .vas a higher TABLE I Sourc@s of Fundkq arad Outmmes of 107 Ciinical Trta(s by Journal ~ ~ r ~En.glxr'd Jovrnal of Medicine Cs s- 1 ~ tazxC e i PS ~ GS ~ , Mn~1s of lnterna/ McYlklne !( PS C.S Favorinu Naw Tmtment Favaina Traditlonal Trsatttwnt ToLaf ~ 1 to 9 8 17 8 0 8 Zz 10 32 6 1 7 6 1 7 3 2 5 2 5 7 7 0 7 4 3 7 Ti1fi.E 2 Sarnple Size psr Grow Rt mumamtk* Slspportsd and Qrsrs{!y Suppostsd CHna! Trtsls• 157 Sarrrpf. stae p.r c.rmip Phwmaceutic.t S4port (%) (n = 37) GWWat 9upport (%) (n - 70) s15 6(16) 15 (21) 1 t~r5o 18(49) 28(40) >50 13(35) 27(39) •No siQNncant Offirance betwem Pnam,acastsotly suppoctna and generaity wppattaa trtaLL proportion of published studies favoring nevw ther- apies funded by drug companies compared with those getting general support. Whilo it is possible that editors and reviewers preferentially rtjsct negative pharmaiceuticallT supported studies, an alternative (and more likely) explanation is ::iat fewer of these studies are submitted. DruQ companms may select foc study drugs that have been previously shown to be efficacious, in some cases in other eountries; these trials aze more likely to favor tha therapy. As a study pnoQresses, and the accumulatinQ resuits appear negatie., companies may conserve funds by discontinuing the study prior to exclusion of a Type II error. In- vestigators may fear discontinued funding if neg- ative trials are submitted and published. and may be pressured not to submit such trials. The com- bination of these occurrences could contsibute to the high publication rate of pharmaceutically sup- ported studies favoring new rather than traditional therapies. While it seems unlikely that conspiracies to suppress unfavorable results of clinical trials exist, a de facto exclusion of negative results may be occurring. There are almost certainly many causes for this exclusion. related to decisions by phar- maceutical firms, investigators, and editors. The characterization of the factors leading to publica- tion has been recently addressed by Shapiro. who noted that "To publish on drug effects is to risk the anger or pleasure of ... practicing physicians, the research community, regulatory agencies. drug manufacturers, and the public. Each constituency has its own unique interests, and consciously or unconsciously exerts its own unique pres.eures.",0 Further investigation into the fate of phar- maceutically supported studies would be aa ap- propriate next step. Research divisions of phar- maceutical firms could provide information regarding the numbers of studies funded and com- pleted, results, and publication outcomes. Critical review and analysia of thia type of informatioa could help ensure the objectivity of scientific research. PS = study whdry or partialfy funded by a pharsnaceutkal firm. The awtor xknwrledges the uig vrxe of 5opfiia Smlth in dve pcrforrtvnce of this GS = generai support (no evidence of support by a pharmaceutical study. George Granasas in the review of the manusrs(pL and Evelyn Lea in the km)- preparaGon of rhe nunuxript.

ISa DavioWn. SouRa OF Furvanc A+w STUOr OuRa.+e REFERENCES 1. Fktcher RH. Fkutxr SW. Clinical rsrrch in general medEcal jourtuls. N Engl J Med 1979301:1t90-3 2 Kw+ln CM. cUnica! (nvestigamn and the pharrtaci,tical kidusuy. Ann lnoern Med 1978;89 (5 Pt 2 Suppi):842-5 3. 1Musher DM. Antffitobcs: a)e medtum Is the message. Rev Infect Dls 19m5:809-12 4. 1(sss Ei. Seductlon in a gcand hotel. Rev Infect Dis 19fl3~5~73-4. S. Kwtin CM. 7?ie tn/atfon b.twan dinkai anvestfgatars and the phar- maceuticii industry. fiev Infect Dis 1984;6:1?9-31 6. Soffer A. ]tazards in puDlkation of proceedings. Atth Intern Med 1982:1422074-6 7. tiodtm's PG. Seminan for physkiatn JAMA 1982Z48:1580-1 H. Ralman AS. Dealing with conflicts of int=rest. N Engi J Med 19®4310:1182-3 9. Ftakmm AS. Daitirg wftfi oonnicar of interest (correspordence). N Engl J Med 1984311 A05 10. Sprung CL CaraHs Pv. Msr<ial EH. et al. Tt* etlects of hlgh-dose corticosteroids in patients with septic shock. N Engi J Med 1984311:1137-43 . 11. Paradst JL, 8lueatipne CD. 8achman RZ, et al . Efefacy of tonsdl.chomy for recttrrerit tfxnat tnfetyion in sevedy aA`e"d dilld<en. ResWts of piraNel ran0onrized and nonrandomlzed tlinical trials. N Ecql J Med 1984310S74493 12. Scfetvttierapy aRer Rrst variceat hamorrfisge in drrtnafs. A randomized mulGc.ntar trtil. Ttti. caputrngm 1;opRag.r vacioia Sdecvthrrapy Ptqiett N Engi J Med 1984311:1 SG4-a70 13. Wtwrw+q PJ. Prfor A. Facsplrr E8. ControtUd trfal of hyrpn~ in tt,e tr..l:,tint of swa. rFtracta,, irtltabte-ttowd syndrome.U,c,i 1 W:Z1 ZM-d 14. Ca~de R. Mars 0. Waprmakec H. et al. Eff.ct of M-obic sxEr dse trair*q on padeMs with ay9zmic artsr{ai hypertmsfon. Am J Med 1g84:?7:715- 90 15. Bow EJ. Louie TJ. RRun P0. Md,augtrtnn RU. Ftxdng GK Ron* AR. Raneortrzed arkr»Hed trtas compartrg tstrn.tt,oprvnrwCa methomoia and trim.trioprim for intLc-tEan prnptt~Aaxis in hoapitAfmid granulacStoPerk padists• Am J M.d 1984:7Yf-??3-33 16. 8uoa JP. Hunn{ngtsaFr Otl. Frantz ID. 1(U» K Mart+sfi CN. Oppel- hNmer JH. Can"rl+on of effecdWenes of thyraoropir~ suEpre~se dosa of dL and 1-thryains In tr.ftmant of hypertta*toeroiemia. Am J Med 19f34:T7:475-81 17. HoMo~ JJ. Smitti CR. Moote RD. FetoU ER, lJ.tman P§. Compafatiw cast etfkllvtttsss of genbmicirt and tabcanydn. Ann !nt=rn Nad 1984:101:764-69 18. Peto R. Plke MC. Armkage P. ac a1. Daa4pn tnd anatys}s of randomiud clinical trials rsquiring probnged a6servatlon of each patienC & J Catxer 197834583-612 19. Freiman JA. Ctykrnrs TC. Smith H. Kuebkr RR. The irtiportance of beta. t3u Type If eirar and aampie alae In the des{p and htxpretation of the randomized controt trial. N Engi J Med 1 f1•18;7.981;90-4 20. Slvipiro S. The deci -:.,n tn pubiish. Etf>Kai diemmss. J Quan Dls 198538365-72 REFLECTIONS [nowl.dQv oad Wisdom, fm from beinQ one, Eiavn ofttims no conn.xion. KnowledQe dw.lls In haads r.rpiet..vith thoughts of oth.c men; Wisdom in minds attentive to their own. Knowledge is proud that he has lamn'd so much; Wisdom is humble that he knows no maz.. Wu.uas~ CoWPEt "Th. Task" copied from Poeticd WorJri, voL II, Little. Brown, Bo.ton. 1859 C a 'A 1< TFiE liP bulato ganim plan fc achm gvcduc demce to som m.d.icc sis of I PI I 1 aa pati m.nbs . whe:e aion of Th aadaF isc not sicianuied3oQ shown nQtL1LB With gI comes tw+..rn actaal tocT es 10 LJ1 O