Tobacco Documents Online

933 presscd. er.{>eciaily in the area of tobacco and health- This article was submitted to the Journal of ClhdcoJ Epldemiolagy becwsse I thaght this epldetnioingical jnucnal wnuld be the <x.c most likely to have an open mind" on a subject as es- cahlished as smoking ets~ation and morplity. After an ex- censive proces6 of reviews, refattions, and revisitsns, the er- ruac was finally accepted r,ttly if It eotdd be accompanied by a"rebuttal." lhe teatt ie now bcirw publlshed as u PR& S6NTATION that is followed by the strong DISSENT of a protninenc efriderniologisr who, withnur disputing the accu- racy of the findings. concludes that it "provides little useful infortnation." It has taken 3 ycars from first submission to ~ennml publication. undergoing a review.process that would thoroughly discourage most epidemiologists from puhli<hinlr such un arrielc or from even invcstig.tcing rhia suhicz:[. In a riQxonfi eXperimeatal science such as physics, facte tltar do noc fit the existing laws of ttature are taken very seriously. Indeed, the observation that Newton's classical laws uf motion were rtvt accurAte ac velocitics approaching the speed of liKhc led to Einsteins special tbeory of relaciv- icy. hven though the relativistic correction is vanishlnGly small for nrdinary terrestrial phenomena, relativity theory is considcrcd cxtrcmely impurrsn[ for the full underst9nding of the laws of natutc- In epidemiology, faccs about subjecu . like eesaacitm that do nor fit dia prevailing are ntn considered valuable or worthy of funhcr invastigation but, r,tthcr, ate dismissed by a technique such as changing thc de(lnitian.osn cessation. Tlris whole process makes me wonder how many itnfwr- a,nt epidemiological issues have not heen fully investiguted nnd/or have bean subject to }n.blic8tion hias. Unless epide- suiologists are mate willinp, t.oallow an open ptts cntacion of xll legitimate findings in an arca as impoetant xs enu.king eessation, parricularly when the findings uta bascd rnt pub- lidy acaileblc data sets that cBn he readily ver!!i , then I think epidcrttioloqy ix in for more of the srrong ritici,m rhar it has faced in recent ycan [1(J-12]. References I. The Hualth Beneflts of Smoking Cemni,xt--A R nf nc, Sur>;eon Gtmeraf. RMd;'i13e, M[? Jirhlic Heahh Service. 199CL DHHS Pvbllcntion No. (COC) 40-8416. 2- Roso CA, Hamilton FJS, C:alareli L, tihipley M). ,smixed controlled trial of anu-amoking advice: Ten-year res ha. J Ey,- idumiol Cmnmunity He.hh 1982; 36: 103-! O8- 3. The MR81T Reacareh Group. Multiple Alsi Factor nterven- ram Trlei (MftF1T) rt.lc factor changcs wnd morrali resulrs- JAMA 1982c 248: 1465-1417. 4. Sharen f.iS. Kuiler LFt. Kfelaberg MO, Smmier J- kene JK, Cudcr jA. cr ¢I. For the MRFIT Reslarch G.wp. r; cnnccr mortality after 16 {rara in MRFIT pmricipar.m In inr tvenrian and usual-cme groufK Ann Eiilcv:n1199; 7: $ 2 5-6. 5. Artthoni.en NR, t.;onnecc JE, Kiley JP. Altoae M . Bailey WC, Bwa AS. rr d. Effects of vrtuekinp inrerventit and the use of an inhaled ancichoiinerRic bronchodihror the uee oFdecline of FEVI (Thc Lung Hmld, Study). J 1994; 272:1497-1505. 6. Ebtnhim S, Smith GD. Systematic rr•ticw vf tands cuo- nulled triklv /d' mulriple risk facu>r interventiWU f prevenr- ingcoronaryhearrdiseace.13b1J 1997; 314= IE6r>-1G 4. 7. Suseer M- The tribulstinnx of t-riaL.-`intervcnitnns i commu- nities fedirorinil. Acn I Public lie.lrh 1995: AS: 15 156. fi. Lillenfdd AM, lilicrdeld DE. Fuundetioas af inlogy. New York: C1xfotd t lnlvetary prca.:1980: pp.10-11 9. Eturmm JE. SmukisrX cessarisNt and mortali:F tmn samong two United Stare< populations. )'C:1in l:pidemiul 999; 52: 813-825. 10. Fumenw M. Sciemr Under Siege. New York: Will m Mur- roar, 1993. 11. Tauhea G. Epidcntkrl.>l;y faces ia limits. Science 1. 5; 2ti9: 1 fi4-159 and EPideme.logy M.nitor, Jane 1996. 12, Sullum J. For Your Own Gacd. New Yu.k: The ' cc Nux; I99H.

J taln Fi`f~km`d vn1. 52. Nu. 9, tT. 831-2131. 1940 t,i~,Yr,rhr C 1799 Phrwicr Scienr.d Inc All ripl,rs rivcrved. VARIANCE, AND DISSENT Re.portae Response to Kullcr's Dissent Jamcs E- Enstrmn St`n<>VLOrPIIPUc'f1EALrn.UNIVEPSRY,,,rC~aIIPOANU,1-oyANGL Y,(;(; Th. DISSPNT dues not idenlify my spe.ciflc crrore in my ,,,,rural experiment methodology or in my findinKs. but nuher questions the moanirig of rhese findingc. Indeed, a mmjjlxv of Kul[er'x Pointc offer t¢rvtible exp(:rnations for my F,tdinKs reg:ardinQ d.e hopt+ct of runoking ccssatlou- AL though I agrcc with much of the DISSENT, there atc scv- enl statcmenes thac ate lnueeamte and(or Indicate a lack of ul.derstandittg of ray 8ndtngs• l provide clarificution here wirh the hope that Kuller and the reader will correctly un- dcrstwnd my findtrtgs and their significance. Because the cme putpuses of my article cr.cnt unclear rn Kulha, let me aitcrate them: to descrihe u potentially seri- ous bia.v in the standard observational epidemiologic+l dare ar+ the value of smoking cessacfott; to meaw.r< the long- trmt popularion impact of cea<adon on two ,:olhurts of older smokers by usirtg a nactaal experiment approach; and to note cfse failure of the intarvention trials involving cessa- nct+ rn crmfitm the value of cessation. The reason my article is important is thac chere is a large dltpacicy between the widely proclaimed value uf smoking cessation and the actual value as measured by intervention triuls and natural euperiments. There is no better ecampk of this disparity than the 1990RL-port of the Surgavn Gen- eral on The Heelth Benefits nf Smoking C.eesacion IIl, pre- sumably thc moet defuticive gcwcrnmenc document on this sobjece. 7lre tepotc begins with summary scatLments from rhe tnp fotu U.S. government health officiala, the Secretary and P.ssistant Seaecary lor Health of the Department of Health and Human Services, the Surgeon General, and the fJireccor of the frentets for L7sease Conem[, ull of urho,n conclude that "sumking eessneionn hac major and immedlatc healrh brneFite for men and women of all ages" [1 1- Thic sweeping aratemetu_apparetttly come3 directly from the MAJOR CONCLUSIONS to the full refrurt- Only in chc body of the report is this statement qualified and the 'elue of cessation more accurately described. Among the qualificattons is the presentarfon of the existing randntnired fntervention trials, similar to uay preaenration, which con- Arklrest eetrrs{,ondarrie m: Jamea V Enmum. P1,.G., M-P.ri. Schrn.i of Pi.blre Hcahh, Uni<•<rsay ef CHJ,wiu, Lnr Ansdrn. CA vOM5-I772. Acccrad 4 Mach 1999. eludes rhar-il ia "irnpossible to rcuch conclusi alx,ur the benefi[s of untoking cesaution from rhese stud' alone ...r• [1]. Howcvcr, the preaencation of the incerv dvn trials• tht gold soutdanS vf cpidemiological merhodu gy, is lim- ired to only about 10 pages spread rhroughout a 628-page report. Also, ubservatiannl dara are presented all causes and for lung cancer cluc show death ratea for fr arnnk- ers ate actttalty higher than for eutrent xtnok during che first I-3 years following eessdtion, conrracilctin the nMirm that there are immediate henlth benefits from e tiem. To evaluate rignmnrsly the populacion imp t of cersa- clon, it is neccs,w,ry to coinperc an intervwui grqip of amoken ehac undergoes a smoking ceuation - aerverition with an appropriacc control Kroup. It Is nor fficitnt to base the value of cessation an sel(-edectcd lotlg-term fnrmer sttwkers: This has been an3 still is theiuncde be- hind the Whitehall Study (21, the Multiple isk Factor intervenrion Trial (MRI:IT) [3,41, and the 4 R Healdh Study (S1. In comparison with observacional diea, the intervention trial design offers rhe potential fo eliminar- ing or more cightly eontrollitiQ bias From the , ection of aUbjccrs and from confounding. This imprsrtan difference was upirarcnc within MRFIT itxlf [3,41: )irlten lyzed as u trial, tht intervenrion and control groups sho red no sig- niticant mortality difterrnces, bur when analyxd un rsbser- vational anrdy, th>se sclFxclecced amokets whrs qui after 1 or 2 yenn generaily had lower death rates than t[t who cs.n- tinucd to stuoke. The inconcluaive resulrs of the interventi trials to darc, lorh among individuals and communities (6,71, and tkx unlikelihrmd of nny furcher major cesaarivn ls, mean chet there In apccial value in ltarutal expetimenrs Involving ccssnemn (8). If a cohon of smokers (argcly lra. their smoking-relaccd death races should eventua[ly verge to- ward thr corresPondirtg death rates amvng the><ce o have never smoked, if ccssation has a ttet posit(ve bta t no the smokers as a whole. Kullcr points our chae the failure to note a ge in relative risk for all causes might be partially dua changcvr in ndter risk facrore that could negtrtc the ben ts of cts- sacion, hot this considararion dnes not apply u the rela- dve rUk fur lung canc~r hecnusc clgarene .emo ' is hy far I-1

832 the strangest risk f,ictor. If there are leccnra that negate the bene9es of cessation, chen these should be identified and the (imitations on cessation noted. Fttrtltermore, I have not assumed char dmokinR cessation does rrat reduce the excess relative risk, but rather observed it through tny resulrs. Ta suppnrr the revtlts nf.mvtwo natu.ral e:pcritnents. I have presenced in Table 9 a caotparison of death rates dar- lnt! 1966-19C8 and 1986 for cdl U.S, whites aged 45-64 yet,rs as of 1966-7 yfi8 that shows no convergax:e of the relative ratio of dcath tyces far the total population atld ncvcr-stnnkcr.r. in spim of substantial ces8atiun in this birrh cnhurt 191. This comparison of cross-sectional sun•eys is rca- scNtebly crnwisnmr with the rwo pro.apctxive natural expcri- nunrs. In addition, my own findings indicate char the uital aKc-adjusted death mte has deelined in the Unired Sraces fvr both smokcrs and twnsmokers from 19G6co 1986 193, but these data are not direcrly relevant to my cohort find- ittgs. The large decline in the cotal death rate among twn- smokcts is nbviously twt related to smoking and indicarea thac all major detnrtninants of mortality trends are nor yet well undetss tood. Kulicr has tni>charactcrited tny s;tacements regarding lung canttr death rate tmnds. The focux of my article is on older lxtscros, and I have presented no data on crends wnwng ye^ut8er persons. Alehwghdeath rarca are declining in younger age gcoupa, snooking trends I have seen suggest that these declines are due primarily to al;mwing propor- ricm of nevet-stookets in ehcse agc groups and not to ea.et- cion. Raccs are nor deciinittx in older age gn+ups, in which maat lutg ca+uxr dcaths occur and Itl which aubsrantial ccs- satiun has occurred. On the basis of the data presented or sutnmarited in this article, I strongly disagrce with the no- tion rhatctiurenclungcat.cerdeachratesate.umsistentwidt staaking tttztds. Keep in mind that the total age-ndfusted lung cancer death rate is at a record high for woman and only slightly less chmt a record high for tnen in spite of the facc rhat per capita cigarecte consutnption, which has i,ecnde- clining for 35 yearx, ts at itr lowest level in SS years- Contrary to Kullcr's contention, I have made absolutely no presutttption that the individuals in these cnhotts stopped smdcing as a good hcalth behavior. Futthetmotc, I have actcmpced to access the full polwlation impact of ces- sation, not just the itnpact attrong healthy indivfu;uals. The ccssacion presumed is all types of cessation, thee seme as for 4 of 5 prospective and 10 of 10 case-wntrol studiea that sumttu.irre the beneHts of cessation for lung eancer in the Suegcon Gencral's Report 111. Thc results for theae 14 snd- Iea were presented for all funtus smdcets without regard to che reason they quit or their health starus. Indeed, only the cea+ks for the U.3, veteraas cohort, thic vcry study, were limited m forma cigarette smokers who quit far reesons other than doctols orders, and these results are essentially identical to those presented in Tabic 2 fix ailformer sewkers. Kulkr Is applying a double standard if he acceprs without qualification the existing observational tpidemiological J. E. Enscrc,m daca that the death races of sclf-selecr f<.rmer stnokers converge toward those of nevcr-smnkers nd then distnlsse:: the findittgs of my natural experlments stdokictg cec.a- timt because I have not detccmined the ns WhY the in- dividuals stopped xmoking. Asse,rsitqt int y those who quit smoking ac a>;cx,d hcalch behavior ia a c and iu=ver hax becn the accepted way that the Iwpac of ezs.sacionhas been evaluated in other st.udietand doe: not.'mestsure the pnl+uletitm ftnt+ect of eessatkm ;?er se. - Kulier hinlself resoFmir.ed dmc the be.ne ts of ctasation as tncasured by cohort and .:asc-coatrol sc 'es aK- subject co limiratiotu, including sr.l4.xlecrinn and misetxollectiun, when he discus,se.d this isauc reeenrly in n a+nalysis of IuaK eanr.rr races in MRFIT 141- hlia analysi eoneludo that "chen:is no evidcnce that thc intetven n lowered luctg cancermtnrality....-'141tatdthatMRFi T asnotprovided dcfinitive eAulrnee for.intervention ben ts for all causes of death either 13.41• Fnr the record, assessed the in- te.tventinn group as a whole and "qutirt were not asked their reason for eessat3tm" 141. Althcntg}t I o nor ktrow the reasons why individuals stopped.smaking have presenccd in Table 8 the key health charactaristicsthe 1971-1975 smolreix in the Nl-IEFS cohorc by their 19 -1984 eitwking smtus and the profiles of eontinuictp amu atidquitters are quite similar (yi- . Kuller's 'statetttents as to why the reduc on in lung can- cer and coronsry heart disease risk foll ing cesaation miKh[ he modest amoctl.' older snwkas quite rcason- able. Also, it may well be that the bertefi of cessttion are, limited primarily to thnse who quit sm'ng as a gcnxd health behavior. One-wnnders why .. qualificaci<mv about the vahu- of cessation, whic{t cto support my findings, are not more prominently irated into the body and cxmchuions of the SurKctm Ge t's Fteport. While I agree that there is solid evid regarding ther cffect;< of cigarette smnking on mortality, do nor' i/vttk there is "equally pow~erftll evuience uE the a of smoking cessacion_" Considering the relacively low rrent smoking prevalence, the continuing high lung eaa er death raeez, the failurc of the Incervention tziala ro c ~rm the benefic of cesdativn, and the curnnt .esults nf t+aw 1-e~,crimtK.ts involving cesariun, ic seems ciear to nw thete art.liut- itations to the health ~Fcacfin of smokinII s• ticm. In conclusion, my study pttppuru ihe I-known find- ings chet the best way to avoid the excess liry esswci- aced with smoking Is to never smokc and the ttoct best way is tn be a long-tenn former smoker. Ii et, ncn all olnukers wlw quic become long-term form tmokenrn. This study indicates rhat the population impact crssariert i.n elderly smokers may be subaantially less i.svridely be=- lieved' and it does tusr support thc noricus e"smoking cessacion has major and itstmediaea}tealch neFta fiu men ~ and women of all ages-" ' CD Finaliy, I would like to point <mc how di ir it is get an UI nlcemativ< intcrpretatiun nf epidemlol cal dara ex- 4M CR tD t0 OD