Tobacco Documents Online

O~ONZ9890Z CHAPTER Keynote Address: The Control of Lung C¢ ncer t RICHARD PETO and RIC HARD DOLL Cancer Studies Unit, Radcliff, Infirmary. Oxford OX2 6HE, Great Britain ICRF Cancer Epidemlo|ogy tnd Clinical Trials Unit, Radcliffe Infirmary. Oxford OX2 6lIE, i ;rear Brilain ABSTRACT During the 1980s, alxmt 1 million Americans anti 2 m Ilion European.'~ are likely to die or tobacco-induced lung cancer, and there is increasingly' h.eaw marketing of manufactured cigarettes in developing countries. This will produce large increases in lung cancer in the next century. These increases are inevitable, just as the increases in United States cigarette usage 40 and more mars ago are responsible for the la:'ge increases in United States lung cancer rates t~May. This increased usage overwhelms the beneficial effects of cigarette tar level reductions. By contt ast, in Britain and Finland mat, lung cancer rates in middle age had already stabilized Ix'fore the large tar reductions begat , and in early middle age the maid lung cancer death rates have already halved, and are s ill dropping fast. Unless recent tar reductions have implausibly large adverse effects on v t~ular or respiratory disease, they have perhaps bccn underrated as practical public he.dth measnrcs. (United States and United Kingdom tar reductions still can be accelerated, md at present many other countries lag well behind.) Although the control of fossil fuel combustion product ~, ionizing radiations, asbestos, and other occupational factors is certainly worthwhile, it can tot avoid any substantial fraction of lung cancer deaths, and although substantially protc~ live nutritional modifications may exist, none has yet been reliably identified. Meanwhile. recent European experience shows that govern,ncnts can, if they wish, achieve substantial ~ cductions in both tar deliveries and cigarette nsagc without materially affecting personal fr 'edoms, and nnless one or both of tbese arc achieved in many different countries there is lltlle prospect of avoiding any substan- tial fraction of the many millions of lung cancer deaths hat threaten to occur over the next few decades. Kqy Words: Lung cancer prevention, cigarette sales tr rods, cigarette tar reduction, lung cancer trends, smoking duration," quantitative inforn~ atlon, clgarette-lung cancer inter- national cnrrelation, cigarette taxation, tobacco prlc" increases 'This article is expanded [mm a report prepared for the WHO ~ancer Unit, b.t it does not necessarily reflect the views of the WI tO or of its Cancer Unit. ~' t~8~ Vedog Cher'n~ Intemot~. Inc. L~ Conce~ C~uses acid Pa~,,,entlon.

2 Rtchord Peto and Rk~hard For the control of each type of cancer, three strategies are available--prevention, screening (early detection), and treatment. Over the past few decades cancer research has produced (or suggested) some important results in all three areas. These include, for example, the effects of hepatitis B virus on the risk of liver cancer, of lower tar cigarettes on the risk of lung cancer, of screening on the prog- nosis of cervical cancer, and of cytotoxic drugs on the treatment of Hodgkin disea~. It therefore is reasonable to be optimistic about, but impossible to predict reliably, what cancer research will produce over the next few decades. If, however, we consider the promise not of what future research may one day offer but of what present-day knowledge can already offer, then the most prac- ticable, and cost-effectlve, opportunities for avoiding premature death from cancer probably involve not screening or improved treatment, but prevention, and this is particularly true of lung cancer. This conclusion does not involve the unrealistic assumption that tobacco can be eliminated: instead, it merely assumes that cigarette sales can be somewhat reduced (eg, by politically realistic price changes, or by the type of education that already appears to have had a substantial effect on white-collar cigarette usage), that the tar delivery per cigarette can likewise be somewhat reduced, and that gro~'~ occupa- tional exposures (eg, to levels of asbestos far higher than are nowadays permissible in, for example, the United States) can be avoided. Nor does this conclusion in- volve the assumption that lung cancer is already a major health problem every- where, for it applies not only in countries where cigarette smoking has been widespread for decades (eg, the United States, where lung cancer already accounts for some 25-30% of all cancer deaths) but also in countries in which cigarette smoking has become widespread only in recent decades (eg, China, where lung cancer as yet accounts for only al:x~ut 5-10% of all cancer deaths). This is because the main rise in lung cancer produced by cigarettes may take as much as halfa cen- tury to materialize, so countries where cigarette smoking is only now becoming or has only recently become widespread can expect large increases in lung cancer dur- ing the 1990s or early in the next century unless effective actiou against the health effects of tobacco can be achieved. The reasons that the prevention of lung cancer is of such overwhelming importance are, first, that the disease is extremely common~; second, dmt it is el'he types of cancer that cause most deaths worldwide are cancers of the lung and of the stomach. Re- cent International Agency for Re~earch on Gaacer/~/HO joint estimates (1) maggest that by 1975 the annual number of new ca~s of lung cancer was already about 0.6 million (developed "west," 0.3 million; developed "east," 0.1 million; China and rest of "third" world, 0.2 million), a total similar to that for stomach cancer. For both diseases, about 90% of affected patients are likely to die. However, whereas in m~t countries stomach cancer rates are either steady or decreasing, lung cancer rates are (with a few interest{ng exceptions that may result from cigarette tar delivery reductions) either steady or increasing. ~, during the 1980s lung cancer will account for more deaths than any other tyl~ of cancer; indeed, the annual number nf lung cancer deaths is pro]mbly already, in the early 19B0~, alw~ut two- thirds of a million, and it may well exceed a million by Ihe end of the century. SEven with expensive hlgh-technotogy support, current surglcM management of lung cancer cures only about 10% of all case~, and ahbough thi~ percentage is snllic~eot to justify the widespread use of surgery (at least in developed countries), it i~ small, has sh~r~wn little change ira recent decades, and is hardly ha- proved at all by the addition of any other current cremes of therapy. l~o~Eg~90e The Control of Lung Can :e~ 3 generally incurableS; third, that effective preventiv ~" measures are already reliably known~; and last, that implementation of these me ~sures will also have a substan- tial impact on many other diseasesL Consequently, what will follow is a description not of early detection or treat- ment, but only of the preventability of lung cancer where chief emphasis must evltably be on the effects of smoking, either alon.: or in combination with other causative factors. There already have been many cxcellent reviews of the effects of tobacco on lung cancer (and, of course, on many ,,thee diseases), and of the pros- poets of controlling the smoking epidemic (:2). The~ efore, the present text is intend- ed not as a balanced (and repetitious) account ef the whole problem, but as a discussion centering on those particular aspects of the relationship between smok- ing and lung cancer that commonly engender iml,orlant misunderstandings, and of those particular aspects of prevention that corn nonly are underemphasized. For a comprehensive review of the overall heath effects of tobacco, the U.S. Surgeon-Gener,'d's 1979 report (3) may be consult.'d [although for a description of the effects of tobacco just on cancer the Surgeon-G ;neral's 1982 report (4) is much to be preferredI. For a review of the overall epidt Iniology of cancer in developed countries, the report of Doll and Peto (5) may be c~,nsulted, especially as Appendix E reviews in some detail the effects of past chanFes in tobacco usage and recent changes in cigarette tar deliveries on lung cancer trt ads. Finally, for reviews of prac- tical steps toward the avoidance of smoking, in addition to the United States Surgeon-General's 1979 report (3), the UICC (6), World Health Organization (2) and Ontario Council on Health (7) reports are im ;duable. Epldemlology of Smoking a~ld Lung Cancer The Need for Prolonge(I Exposure There are a few key features of the effects of 'trbacco on lung cancer that are slightly counterintuitive, and these are discussed ::t some length by Doll and Peto (5). Ohlef among them, and the key to any prol~ r understanding of tobacco car- clnogencsis, is the extraordinary relevance of the d tradon of smoking to hmg cancer onset rates. For example, after 45, 30, and 15 year : ofclgarette star,king, the excess annual incidence rates of lung cancer may be about 0.5%, 0.1% and under *The most effective step is to avoid increases, or to produce t rcreases, in cigarette consumption, but changes in the harmfulness of cigarettes also may be]p. There may be consld~erable pelitlcal difficulties in taking any actions that will affect cigarette consumption sub~ tantially, but it is nevertheless likely that ~omc such actions will be practicable in at least some counh ies over the next few ycar~. Moreover, substantial decrea~s in the sales-weighted mean tar deliverie.~ of manufactured cigarett¢~ are likely to be p~litically practicable in many coimtrles, as they have l'ttle economic impact on governments, rl'obaceo is al.~ an important can~ of various less common tyI cs of cancer, eg, of the mouth, pharynx, larynx, aml e~phagus, and p~bably aim of the panc~m~ and trina~ tract. Mo~ im~flantly, at least in devrlol,d cmmtries sm'h as Britain, to~c~ p~bly ]ills more ~ople from ~splrato~ and va~ulnr di~a~ than irma cancer.

l~chard Pelo and l~hard Doll 0.01%, respectively (Table I), The annual lung cancer incidence rates to be ex- pected among smokers may be. estimated by adding up a background (nonsmoker) rate, which, like the onset rates of many other types of cancer, depends strongly on age (but not, of course, on tobacco exposure), plus an excess rate, which depends Table 1. Approximate" effects of various durations of cigarette smoking on annual incidence of hmg cancer Annual excess incidence Years of Moderate Heavy cigarette smok|ng smokers smokers % 15 0.005 0.01 30 0.1 0.2 45 0.5 1 (60) (1.57) (3?) aF, stimated from data reported by Doll and Peto (8) for male British doctors. The cumulative risks would be far greater than the.~, annual risks, of course, so an eventual total of over 10% of reg.lar cigarette smokers may die of tnba¢co-induced cancer, depending on the number and type of clgaretles smoked. Annual lung cancer death rate per 100,0OO men (standardized for amount smoked) 1000 100 10 Smoked cigarettes ~lnce age . _ ,~m Never ~" - smoked - / regularly 20 40 60 80 Duration of exposure in years (age - 221/2 for smokers age - 2th for nonsmokers) Figure I. Background and excess risks: lung cancer death rates autong (a) non-smokers (lower line) in relation to age, and (b) regular cigarette smokers (upper line) in relation to ap- proximate years ofsmoking. From Ref. (9). These two lines can be used directly to indicate the approximate baekgrmmd and excess risks, for in middle and old age the lung cancer in- cidence rates anmng people who have snmked cigarettes tt~r.ughuut aduh life greatly cxrccd the rates among nonsmokers of similar age. (This might not, however, be. true for t~oplc who did not I~gin to smoke substantial nmnbe.rs of cigarettes until middle age, for the background and the excess annual risks per 10O,000 men indicated by these lines are, rcspcc- lively, approximately IOs times the fourth power of years of age 10"4 times the fourth power times years of regular cigarette smoking.) g~O~g9890E The Conlrol of Lung Car 0.3% 0.2% O.1% 0.0% ~Y Before 15 15-19 20-24 25or Never o/er Age (years) when s arted to smoke cigarettes Figure 2. Tim relevance uf star,king iu early aduh life: , lationsbip, in prospective survey data of regular smokers, between the age when regular 'igarette smoking began in early aduh life and hmg cancer deatt~ rates at age 55-64 (mean * age 60) for Unitod States males. From Ref. (,5). Data are presented separately for heavy a td for moderate smokers. strong!y on duration of regular tobacco exposure (I ut not otherwise, at least to a first approxi~nation, on age). Typical background ;md excess rates for males are depicted in Figure 1 (9), and those for females ma5 be about two-thirds as great. The most surprising consequence of the overwl'elming effects of the duration of smoking is illustrated, using real datas, in Figur,. 2, which shows how strongly the annual excess risk of death from lung cancer ;~t 60 years of age depends on whetber men started smoking at 15 or at 25 years of age tie, on whether by the age of 60 they had smoked for 45, or for only 35, yt ars). Failure to appreciate the relationship illustrated in Figure 2 has led to a v ~riety of unjustifiable conclu- sions, eg, that cigarettes do not cause lung cancer cr, less perversely, that low-tar cigarettes have at least as great an effect as high-tar ones (10); that air pollution is of comparable imp(~rtance to tobacco [see, however, Cederl6f et al (11)]; or that new on.sos tff hmg cancer (rather than the dela)cd effects of past changes in tobacco usage) are chiefly responsible for the rapid ncreases in lung cancer in re- cent years. In each case the point that often is over'ooked is that current patterns ofhmg cancer mortality rates in late middle age or ,n old age depend strongly not only on current patterns of tobacco usage, but als ~ on the patterns of cigarette usage anlong young adults as mncb as half a cent try ago. t'llw data utilized are frmn the third largest prospective survey ) 't reporled anti are similar to the find- ings in lhe larger two surveys; the corresponding results from all ~hree of these surveys are presented in the II.S. Si.gcon.(;rnr~'al's 1982 trl~tl (.I).

r'dchord Pefo and Rlchord Doll Therefore, current trends, current urban/rural differences, and current inter- national differences in lung cancer reflect, among other things, past Ircnds, past urban/rural differences, and past international differences in cigarette usage by young adults. Consider, for example, the extent to whicb current trends in United States lung cancer mortality rates among men now aged 70 may be affected by the large trends in cigarette consumption 50 years ago among people then aged 20 (5). (For details, see Appendix E of Doll and Peto, (7). In 1930, United States cigarette consumption was increasing rapidly among young men, and national sales rose from 1 cigarette per adult a day in 1915 to about 10 per adult a day in 1945. The effects of those increases are only now becoming fidly apparent, and largely as a very long delayed result of them, United States male lung cancer rates in late mid- dle and old age are still rising steeply, despite the fact that cigarette sales per adult have remained at approximately 10-1'2 a day ever since 194-5, and tbat tar levels per cigarette have fallen substantially (Figure 3). Contrary to various suggestions, the "discrepancy" that has been seen for the past 25 years in the United States be- tween rising lung cancer rates (see Figure 4) and falling tar levels does not imply, or even suggest, that Americans are exposed to increasing levels of carcinogenic pollutants other than tobacco, or, as the recent (10) National Academy of Sciences-National Research Council (NAS-NRC) report suggested, that tar level reductions in cigarettes have been ineffective. Indeed, but for tar-level reductions, the current increases in United States lung cancer mortality rates probalfly wouhl be appreciably more rapid. Likcwi~, in many countries the smoking of manufactured cigarettes by young adults was a habit that tended to become established first in the towns before it spread to the surrounding countryside, rather than the converse. Consequently, 15 Actt~al cigarette consumption per adult 0 ._~ O /- cigarette equivalents) 19~ 1920 1940 1960 1980 Fi~re 3. Trend~ in United ~latc~ consumption: mean daily ~ale~ of mamlrarl,red ciga- rel~c~ l~r Untied Slale~ athlll aged over lfl year~, It~elher with a £rtltle e~timal~ of lar yield l~r atlull. Fr.m Ref. (5). "ltm estimate of tar yield allow~ approximately fi~r tlecl'ea~es ~hlce the 1950s i, lar yield ~r cigare,e sm.ked in a standard maturer, hut not fi)r any syslemalic changes in lhe manner in which cigarettc~ are smoked. The Control of Lung Can :er 7 half a century ago cigarette smoking was probably more prevalent among young men in towns than among young men in the corn try. Disparities in recent years between urban and rural lung cancer rates amon! today's old smokers therefore may rcsuh chicfly not from air pollution but rom a delayed effect of past urban-rural differences in cigarette usage among t ve people who were then young but who are now old. Finally, it is wholly wrong to suggest that the p~ or international correlation be- tween current smoking habits and current lung cmcer rates indicates tbat smoking is not the chief determinant of worldwide hmg ca~,cer mortality. For, such a cor- relation effectively relates lhe lung cancer rates of t ~e grandparents to the smoking habits of their grandchildren. If instead the nati ~nal lung cancer rates for one generation arc related to national cigarette consum! ~don rates when that generation were young adults, a moderately close relationsbi], does emerge (Figure 5). 25,000 20,000 15,000 10,000 6000 2000 0 "Estimated rates among nonsmokers ~ Mouth, esophagi l:lh~ rynx or laryn ~ Figure 4. Recent trends in United States cancer mow Mity rates: age-standardized death ccrlificadow ral~ (Iwr I(~ million i~.~q~le a~cd .rider Ti~ ) in flw 19~Os (lop bar), flw I~ (middle I,ar) aml dw 197(E (t~)ttom bin') G)r vmi,m~ ~ .'~ .f ennecr in lhc Untied From Rcf (5). For cancers of the long avul .])l~'r rcspir~ h,ry and digeslivv tracls, estimatt'd rates for lifelo.g IlOnsl.nkers are alst~ given (asterisked ;m', nlxwc the rates for the 8~O~g9890g !~

r'~hord Peto and I~herd Belglu.m Canada ~ France New Zealan Germajy ~ / b~~Greece " ~ 60 ~pain / ~ ~ ~ . • Rates based on o~r tOO deaths m ~ ~ ~~ Den~ ~a/e~sed on 25-~00 deaths ~ ~o n D Noway 500 1000 1500 2000 25~ 3000 Manufactured clgaretles ~r adult In tg50 F~gure 5. I,un~ crower m~d mno~h)~ hi Ihe ~ame ~enera(ion: rclmi.aship hclween hmg m;malhrl..'d (exrl.dc~ tmndmlled rig~rettes in I~l~ium cigarette rousumpli~m when flint gcncralioa of ~ople were in early adult life: (lala various cou.tries, and for US non-smokers estimated hy fittin~ Che p~s[~ectk'e s.rvey d.¢~ re.reed by dm American Cancer ~'icCy (25). Other Features Dose-Response Relationships In Table I, it may be seen that doubling the dose may approximately double the excess risk at each age. Partly because ofdifficuhies of dosimetry~, it is not rcaUy known whether, as Doll and Pcto have tentatively suggested (8), a doubling of the true dose rate produces an approximately fourfold increase in the age-specific effect, or whether, as is suggested by much other data, it merely prodnces a twofold increase. Whatever the exact truth, however, it is clear that two packs a day for 20 years is far less hazardous than one pack a day for 40 years, so any reports based on inappropriate concepts such as "pack-years" should be treated warily. Effects of Stopping Smoking When smoking ceases, the annual excess risk remains roughly (perhaps to within a factor of two) constant thereafter. Referring to Table 1, it may be seen that the annual excess risk after 30 years of s~noking is about 0.1%, ~ if a smoker stops el'he cffedivc (k~se may .at ~x: simply proportional to Ihe nnmhcr ofcigarc.es sm.ked per (lay, for the CO uptake fx'r cigarette apl~ars to be less fi~r heavy than for mc;~lerate srm~ker.~. AI~, t~.¢ause the chief target area i.~ the main airway~, r~pid inhalatiem may ¢Jcp,'~sit les~ rnl thegn/Iron ~h~w inhalali~m does. 'lhi~ s.ggestl.n j~ rr'~'~tfi;rr'erl by relents (12) that i. ~mw, altlto.gh .at all, ~t.rlies he;try .'o.oker,.~ who describe themselve.~ as "n,r~t inhaling" get m~z~ hmg cancer than do co.q~arably heavy smnke~s wh¢, "do inhale"!. ~o~ggggog The Ca431rol of Lung Can :er 9 after 30 years, then approximately this annual exce :s risk may persist indefinitely. Thus, for example, 15 years later the annual excess ~ isk may still be about 0.1% in- stead of the 0.5% that it would have been had smokh~g continned, so about 80% of the excess risk is being avoided. It is not true, bower :r, that the annual ab~lutc ex- cess risk decreases substantially, and still Icsa is it tn c that it decreases to zero after 10 years; only one prospective study has suggest~ I that, and the others clearly refute it. But, the large increases in risk that would otherwise happen are avoided by stopping stnoking. Tf'~9 Importance of C!gareffes as Opposed fc Pipes In Britain and tile United States, cigarettes app~ ar to have a far greater effect than pipe or cigar tobacco did, and so the switch ea "lier this century from pipes to cigarettes has produced vast increases in lung cat car. The reasons for this dif- ference are not adequately known, especially as the ~moke from pipes and cigars is about as carcinogenic as that from cigarettes for lal ,oratory animals. One sugges- tion is that the difference depends chiefly on the ,t rearer alkalinity of the smoke from pipes and cigars, which may both make inhalalion less pleasant and facilitate the transport of nicotine across the oral mucosa, tht reby obviating the need to in- hale (13, 14). This suggestion may not be difficult o test and, if confirmed, may point to an important way of diminishing the hazat tls of cigarettes, bat at present this remains speculative~. A related suggestion is tit ~t the "air-cured" tobacco of, ft." cxamph', certain French cigarettes ~nncwhat ~cscmblcs pipe tobacco and is therefore substm~lially less carcinogenic than the "quc-curcd" tobacco typical of British aud American cigarettes, but the intcrnation d differences in lung cancer on which this suggestion rests owe so much to differ 'nces in duration of cigarette smoking that it is still unclear whether there are also any material differences in the hazards of the various cigarettes. (During the 1.¢ 30s and 1940s, for example, British cigarette cousmnption was four times that i t France.) Interaction with Other Causallve Factors A variety of other causative factors for lung cancc • are known, of which the best studied are asbestos, ionizing radiations, and urban air pollution. All thcse have a far greater absolute extra effect on smokers than ¢ n nonsmokers (illustraled for asbestos in Table 2), as may various other causative actors. ,Some of the benefits of control of certain other causes o flung cancer thereto c may be attainahle indirectly by reducing tobacco exposure. However, because e ffectlve tobacco exposures are currently increasing in many countries (and even v.here they are decreasing, the immediate decreases are unlikely to be enormous), the theoretical possibility of avoiding tobacco exposure clearly does not justify in action where other substantial causes of lung cancer can be reduced materiallyL qn the stn(ly of Cc¢led~3f el al (15) in Sweden, p~pe ~mokcrs had the same tcnfokt cxces~ of lung cancer that cigarette smokers had, which rather sugges1~ that the sm ,llness of the effects in Britain or the United State.¢ may res.h more from traditions almut h~w pil~ ~ are smoked than frmn the pharma- colr~p/of dw smoke--and it i.~ unlikely that such traditinn~ wil dwmsclves Ix. wholly dcternfincd pharmacoh~gic faclnrs. °Apart fi~...~rm,kh~g, a~l.'stos, kmizlng tmlintlons, aml comb. li.n pr~*dn('t~ .f fi,~il Ihcl~, the reliably cstabli.~hcd can~c.~ nfhmg cancer are. bi~ehloromelhyl)ed'cr (BflME), mu.~tard ga~, and certain comlxr.nds .r oxidation ~tat~ of As, Gr, and Ni (5).

I(9 Richard Peto and Fdchard Doll Table 2. Multiplicative effects of heavy asbestos exposure and of smoking on lung" cancer risks" Relat|,ce risk of I_u.0ag c~a ricer for: Nonsmokers Smokers No known asbestos Heavy asbestos exposure (prolonged employment as a lagger before 1968 United States dust controls were introduced) ! (reference category) 5 aDma rmm Selikoff (16). Note that although such heavy asbestos exposure is no longer permitted in many countries, places where heavy occupational exposures do still occur may offer excellent opportunities for limited disease prevention, because even if the workers do not smoke (=, the excess risk of bronchial carcinoma is low), the risk of me*othelioma, which does not depend on synergy with tobacco, will mill be high. Mlscerllflcatlon of Lung Cancer Deoths People, and especially old people, dying of lung cancer may never have their disease recognized and may be miscertified as dying of ~me other condition. Pro- gressive rectification of such errors produces large, purely artifactual, increases in hntg cancer death certification rates. In middle age such effects were substantial during the first half of the century, even in developed countries--for example, when diagnostic radiology was introduced dtnring the 1920s, it prodnced alxmt a threefold incream in British lung cancer death ce~ification rates--but in midclle age st,oh effects are now limited chiefly to underdeveloped countries. In old age, however, large (eg, twofold) artifactual increases have continued to occur since 1950, even in various devclotx'd countries, whereas among old people in many underdeveloped countries lung cancer death certification rates are still grossly unreliable [as are "age-standardized" lung cancer death certification rates, unless standm~dization is to the truncated age range 35-64 recommended by the Inter- national Agency for Research on Cancer (IARG), (17) to circumvent such difficulties.] Tar Deliveries The effects of changes in tar deliveries need to be properly understood by anyone concerned with the avoidance of cancer, for at least in developed countries they may offer one of the more important cancer control strategies. Between the 1930s and the 1970s there have been reductions of more than 50% in the mean tie, sales- weighted) tar delivery per cigarette in the United States, Britain, Scandinavia, and a few other places. These changes were small until the late 1950s and then they sud- (h'nly l)vvame rapid, with dcrrcases fi'om 30-odd mg per clgarc/te in ei~e mid-1950s down m alqm~ximau'ly 15 mg IWU clgmelle hydtc 1970s. The chau~l{es me n~t ex- pensive, and involve d~e use of fihcrtips, porous paper (or even, as an extreme measuH', "ventilated" fillers 0rot allow air to tiller hllo the side nf the filler t(i The Control of Ltn~ Cancer 11 dilute the smoke) aod modified types of tobacco (which may in some instances ac- tually h,e less expensive than unmodified tobacco). There is, of course, a reduction not truly in the unwanted componenls of the smoke but also in those substances (eg nic'odne?) to which some smokers are adclicted, atttl when snch reductions occur many smokc~s a~ likely to com~nsate, either by smoking more cigarettes~ or, ~rhaps more commonly, by taking in more smoke per cigaretteu. It appears, however, that the latter form of compensation is not Mways su~cient to outweigh the reduction in tar (19), in which c~e the net ~sult will ~ inhMation of less tar into the lung. This conclusion is suggested ~th by common scn~ and by obse~a- tion, but cvcn if it is accepted it does not prove that the h~ards will ~ correspon- dingly reduced, for despite some 30 years of la~ratory resea~h the im~rtantly carcinogenic factors in cigarette smoke have not yet been identified reliably. Moreover, it is di~cult to predict how changed patterns of inhMation will change what is de.sited on the main target areas--which, for lung cancer, are not the peripheral tissues, but in the large airways~ the smoke streams past them. Gonscqucntly, it is necessary to discover by direct epidemiologic observation whether the risks of lung cancer are materially reduced by the widespread switch to lower tar cigarettes. Unfortunately this is not easy to do, for not only are smokers of low-tar brands self-selected but al~, just as it is only a~er some decades of smoking that the full risks matcri~izc, ~ ~rhaps it is only after some decades of using low- tar cigarettes that the fitll benefits will materialize, Therefore, even if the effects in late middle age will one clay be substantial, they may not yet I~. Any substantial effects that arc going to materialize in ca@ middle age should ~ beginning to be evident by now io Brilaln, however, for although the tar rcdnction~ of the 1950s were nnly moderate, lhosc of the 1960s were substantial in Brilain, North America and Scnndinavia. Thus, a 40 ycnr aid in 1980 will have been smoking from about 1960 to 1980, Ihroughoul mosl of which time tar levels were substantially lower than in previous decades. Two main pieces of epidemiologic evidence are currently available, the first being the rcsuhs from classical case-control or pros~ctive surveys. Unfortunately~ such data as are currently available are limited by the fact that they relate chiefly to late middle or old age, when most of the lung cancers occur, and even recent ]sin principle, tar reductkms could either increase or decrease the number of people who smoke (by making it less of an ordeal for nonsmokers to acquire the habit or by making the habit leg'~ addictive) and coukl either increa~ or decease the number of ciga~ttes one iodlvklual smnkrr consumes (by in- creasing the rmmber needed to achieve a given do~ or by decreasing the satisfaction ~r cigareHr}. In practire, lmwcver, the patterns of cigarette consumption in different conntries do not ap~ar to l~ in- flnenced consislenlly in either direction by chants in cigarette consumption. ttSurptisin.gly, there appears to b~ little reliable information on which of the many characteristics of the cigarette (eg, nicotine, draw resistance," taste) im~antly affect "com~nsad~." If these cot]hi be i<h,nlifivd and lntalified {<.g, b~ invreasing the niroline deliver y. draw resislan<'e, or whatever of law-tar l~lii In'e~uoIMIl7 iln~'lliel Ihe reel'nl di~alll~finlloI ~ndloI h7 K~il[rtia~ ~1 al (111) Ihat ill* ri~ of lii7~w~fdilil hifiirrlhlli all" irli lnalerllill7 di~erenl miumt lillt~ker~ of differeni 171~ ill £~'08E9,£908 .., • Shiffman Medical Library, 4325 Brusl~ St,'

12 f'dchord Pefo ond r'dchord Doll studies relate chiefly to people who have smoked low-tar cigarettes for ouly a frac- tion of their smoking lives. This dittqcuhy is exaccrhated in studies perfi~rmcd dur- ing the 1960s (or early 1970s) by the fact that the tar reductions then availablc study wcrc not only more rcccnt, but also less extreme, than those curttally available. A related source of di~culty is that as overall tar levels dccrcasc, the highcr tar levels simply cease to exist, so direct concurrcnt compari~n of people now on low-tar cigarettes can ~ only with ~oplc on re@crate-tar cigarettes, anti not with the old very high tar brands. ~spite the~ di~cnltles, when I~e and Gar- finkcl (20) reviewed MI the case-control and pros~ctivc studies then available they concluded flint: a reasonably clear picture has emerged. 33fis is that smokers of fihcr (or low tar/nicotine) clga~ttes have a lower mortality than smokers of plain (or high tar/nicotlne) cigarettes for tho~ dise~s ~st ~trongly a~iatcd with smok- ing .... 33~cs~ reductions in mortality have l~en ~n in those who have smoked the more modem ty~s of cigarette for only a small part of their smok- ing livc,. "~e fact that those who have smoked them fi~r longer show even grcawr ~cluct~ns in mortMity ~uggests that the ovrrall pictu~ will impure even more in ~ea~ to Com~. Becausc of di~cuIfieB oE sclf-sclccdon, of comparln~ the ncw with thc old c(m- currcndy, and of characlcrizing individuals' recent hmg cancer ralcs in early middle age tic, the rates among people who have smoked low-tar cigarcttcs for much of d,~ir aduh lives), d~c case-control and prospective survey data cau I~ sup- plcmcntcd Uschdly by a second type of cpidcmiologic data, ic, thc stody of nadnual trends in early mkkllc age. However, fi~r reasons that already have ~cn discusscd, it is not advisable to use for this pur~se data (such as those from the Unitcd States) in which any downward trends caused by tar reductions arc likely to ~. diluted or even rcvcr~d by upward trends resulting from the delayed effects of past incrca~s in tobacco consumption. Instead, it is ~tter to use the British data. For by the 1950s (when thc rapid tar decreases began) British mMe lung cancer rates in early middle age had Mrcady approximately stabilized (Table 3). Table 3 also descries their subsequent evolution, and the reductions are extremely impressive. They are most unlikely to result f~m changes in air ~llution, for not only are any effects of air pollution likely to ~ far smdlcr than this (11), but ~so similar hMvings in early middle age have ~en seen over the last 20 years in un~lluted Finland. Moreover, ~th in Finland and in Britain the changes appear, if anything, to ~ accelerating downward, so if this pattern carries on into late middle age during the next decade or two, thcn at least in these two countries (where the male death rates are at pres- ent uniquely high) lung cancer may some day decrease for a few years~: as fast as it once incrcascd. A finM piece of human evidence that tends indirectly to confirm the reality of these changes is provided by a comparison of histologic sections from American "I1 will not de('rea~ to anywhere near non smoker rates, however, unle.~ there is widespread abandon- meat of ciga~tte smoking. ~milarly, in tho~ other ~pulations where lung caner rate~ have n(H y~t completed/heir ri~, even a tar-level ~luction dmt halve~ the c~rcitmgeniclty nf ciga~ttes may merely ~low, rather than reverse, the progressive increaw of the di~ase ~er ~l~e next few decades. 9~OB39B903 The Control of Lung Can, er Table 3. Recta! trrnds in England an,d Wales male h, tg cancer death ccrlificafion rates in early nti(hlh, age"'t' Ikath certification ratea per million ~ ,en from cancer~ of the re~plratory Age 1951-55 1956-t 0 1980 Ratio (r") (") (b) (0 (~b) 30-# 3B~ 37~ 13 0.3 35-9 lOP 95~ 45 0.5 40-~ 253~ 256~ 134 0.5 45-9 58~ 59~ 37~ 0.6 annie bo¢~ tlw approximate constancy heft)re tar deliver|e~ I'x'gan In Iw greally ~t~d and the la~e drr~axe thereafter. bsol~: (I)'[lw~'lrrnd~nlr.~tmalrtiallyaffi'rlrdhy~hml~r~in~l~rali~, I,ralmrnl~,ft[wdi~a~' (2}Sah,~-~.rlghtrd Mran ~ igarrlte t ~mxU~nlaitm Iwr Brilixh male aged ~-50 did n,,I hangr greatly until the pa,I few yrarx and in 1955. H~5 anti 1975 wn~ rr~pr(liw'ly, 10.5.9.9. and 10.2 121 ) Ihe I0*~1% d~'rea~e in (tm~umpti,m thai have likrwi~" }~'t'll appr.M~natrly Iml~ m'er the im~t 20 years. 1 '~d in l~lb n~mlr~ the de~ rra~x apl~ar, ir dlligh inlnke t~llly in fir~ t~ or m~ tff ~noking hi~tory. aulopsics in the 1950s aud in Ihc 1970s (22). In lhc IqS0s smokers I~atl a high, dose- related prevalence of what were thought to be prene q~lastic lesions, whereas by the 1970s such lcsious were an order of magnitude less ~onunon among smokers. The exact biologic significance of lhese lesions, howcw r, remains obscure, especially because their prevalence decreased so sharp/y durin ; a period in which lung cancer rates were rising. (They may be indicators not so n,uch of the extent to which the main neoplastic endpolnts are occuring, but of tl e extent to which one of the "stages" of carcinogenesis is occurring.) Practical Actlo~: Discouraging Sales and Decre~slng Tar Levels Sales A varicly of WltO and UICC expert reports hart. been prepared on how volun- tary organizations and governments qan decrease ci! arette consumption, and these deserve carefid scrutiny for they contain much ~ ::ll-judged advice. There are, however, two important respects in which they m;~v be somewhat deficient. The first is that lax increases may .nnt be suffieientl) emphasized. Because many gtlvt'rnltltqllS drrive large tax yichls fi'om tobacct, sales, all hut this one of the strategies dmt may be considered for reducing cigar,'ttc sales, will also, if effective, r(.ducc Ia× rcv,.nucs. Ahhough in principle governn ents may t~/ieve they act only for Ihe good uf their chizens, in practice they may tend to dccide that what is

F~hord Peto ar~:l FOchord ~ Table 4. Elasth:ity: Predicted change in cigarette sales per 10% increase in real price Country' studied Estimates in 9~ different papers ofthe change in annual cigarette sales associated with a 10% price increase (%) United States - 5, - 8, - 4, - 4 Canada - 7 United Kingdom -6, -5 Switzerland - 8 Finland - 3h aBelween 1950 aml 1968 no studies of the rla,~ti,'ity of cigarelte demand were published, but since 1968 at least ten have Ix.ca. All are cited except G~r thai of Atkinson anti Skcgg (1974), which is s.pcrseded by the reanalysN by rl Peru (t974) of the identical data. For re.re.fences, see Ontario C, ouncil of lleahh (7). This estlmalr of only 3% was published in 1974 and would have b~en mm'e extreme if 11 had I~n ~sslble ~o in- dude the ~ul~eq.cm large incrca~ in prke and decrease in constanp~br~ tha~ ~k place in Finland i~ mid- 197~. cconomically easiest fclr thc government is bcst for thc citizcns. Consequently, the otto strategy--increaslng the tax on tobacco--that increa~s cadger ~han decreases ~ax revc..es l)crhaps deserves more emphasis than it usually ,gels, esl~:ially because it is one of ¢l~c few straWgies for which thcrc is clcar, direct evidence nf ef¢~:ct, In ]gBI, for example, increases by n total nf at~m~ 2(~% in British cignrcltc prices pr¢~duccd decreases that, althoogh substanti~, were Icss than %)% in cigarette sales, so the tobacco manuhcturcrs complained of unemployment in thc industry while tim government collected marc t~, The same thing hap~ncd in Finland in the mid-1970s. Several reports during the past 15 years have examined marc formally the quantitatlvc relationship ~twcen price and consumption in these and various other countries with remarkably consistent findings ('Fable 4). At Ic~t for the subsequent year or two, a 10% increase in price appears to produce a~ut a 5% decrease in consumption. If such a decrease were hrgcly permanent, dmn it would in the king term prevent abont 10,~0 lobacco-induccd deaths per miffion cigarette smokers, I~ is more di~cuh to prmlucc a reliable estimate of the extent to which these year-to-year changes in consumption, produced by price changes, persist over longer ~riods, for so many other hctors also may bc in- valved. Despite this, however, some dircc~ cvidcncc for the common~nsc notion that price does affect long-term, as well as short-term, consumption is afforded by the general tendency for cigarcttc consumption to ~ high in many countries where the price is low (7). [n view of such data, the promlsc ofdclibcratc shifts finm other g~x)ds to tobacco may deserve greater emphasis than it oficn rccclvcs. A second dcfi¢'icncy of emphasis is that lhcre may have bccn insu~cicnt stress on ~hc longJcrm advantages of getting quantitatively infi~rmativc material across abrupt ft.: l~lal risks from tobacco, and tim cxtc.t l~ which, at least in countries, thc~ exceed all other reliably known causes of death. "l~c reasons any ~rious i.~gram of canccr prevention must strcss thc hcalfl~ clTccts of tobacco arc illustrated by Table 5, which has t~cn abstracted from thc chapter nn Cancer Epidcmiology in the Oxford 7~xt~k ~M~icine (~3), This ~rs~ctivc, howcvcr, is Li~O~ggggog The Contro~ of Lung Can :er 15 Table 5. Reliably established, practicable" ways of a, aiding the onset of life-threatening cancer in the United States or United Kingdomb. Percentage of all US/UK cancer "teaths known to be thus avoidable Avoidance of tobacco smoke 30 Avoidance of alcoholic drinks or mouthwashes 3 Avoidance of obesity 2 Regular cervical .'z'reening and genital hygiene 1 Avoidance of incs~ntlat medical use of hormones or < l radlnlogy Avoidance of unnsnal exposure to sunligh~ < I Avoklance of known eff~ts on ~ple of current levels of ex~,snre m carcinngens Occupational context < V F~, waler or urban air < 1 aExcl.dlng ways sm'h as pmphylaclk p~statcclomy, mas~ccto, ~y, hys~ereclomy, ~pho~mmy, arlificial hFmm [~[ a~d P~'¢¢' (23). rThc p~orlion of current United Sla~es cnnrer deaths Ihat are I kc[~ to rcsuh from ~cupal~nal factors was or pas~ exi~,~u~ ~o aslwsl~ may ~c~ ~mnt fiw I-~% ~ all curre, ~ United ~ate~ caner deaths, slill rising and lhnl eve.l.ally may well Iw 2-3%. However. l~ca ,¢ Ihe nl,pmxhna~c magnit.de ¢ff dw health rffi,t Is .f rx[~.~ to ndwslt~s I~amr wJdr}y nvrrplrd, ex~,sure I," cls have Iwen grcally reduced and are now in ca~in,~ns (~, [~n~kfine), the ~la~ cff~'ts M ~dch a~ ~ill apf ~.afing. l~mg after ~tand~ ~t~ in ex- ~ ha~ ~aken pfa~. unfamiliar even {o mos¢ cancer research workers. Jet ~]one to most nonmedic~ ~ople. Indeed, in r~ent surveys in Britain most ~ ~ple mistakenly imagined that {ra~c caused marc deaths than tobacco; in fact [ ~bacco causes over 20 times as many UK dca{hs as trn~c. Likewise, in the Unit~ ~I States recent surveys have in- dicated flint runny people believe that backgroun~[ radiation from nuclear power plants is n greater hcnhh risk than tobacco (24), ~ hereas in fact tobacco is several thousand times more iml~rmnt. Such gross misp,.rccp{ions, of cou~e, may have substnnfial effects on behavior. Indeed, the ch~irm an o~ R.J. Reynolds, America's largest cigarette manufacturer, rc~r¢cdly (New Ye,k Times, April 12, 1981) said to his sh~rcboJders [hat the reason ~he cancer.scare was no longer hitting cigarette sales ~ hard was that so many things have ~en linked to cancer that ~ople were "beginning [o take a more objcctlve [sic] view of the heath evidence"~ He may well I~ right about [he effect of the string of ~"~rts a~ut new carclnogcns, l~causc truly a remarkably l~rciplcn[ newspa~ reader or televiewer wash{ able [o gncss, after rcadlng abou{ ~nc new cancer scare n~ter another, {hat old sou-newsworthy [nbacco was still causing a}~u{ one-tiffed of dcnlhs~nn effect {ca times as large as the next mos¢ im~rmn[ reliably known cffcel. It is admittedly difficult (o comm.nirat~" risks in a way Ihat will be tmdcrstood and rcmcml~rcd approximately cor~ ,'ctly, csl~ciMly by pcoplc who have no framework oft~flmr risks with which to co,aparc them, tlowever, it should ~ ~ssiblc as }ong as the main message is set clcm~ y apart from the lcsscr messages

16 i"dch~rd Peto and l~chard Doll that qualify it and Ihat may help prevent people from rationalizing it awayL~. Aflcr all, the chief message is merely that "ABOUT A QUARTER OF AI,I. REGULAR CIGARETTE SMOKERS WII.,L BE KILLED BEFORE THEIR TIME BY "FILE IIABIT," which is consldcral)ly less complicated than tim mass of quantitative infornmtion about house prices, groceries, car prices, clc, that already has become part of the folklore of consumer societies. How exactly this main men, age should be put over is a matter for experiment; comparisems with other condidtms may (especi,',Jty in Britain) be helpful, eg, SM()KING IS BRITAIN'S BIGGF~'q'I' KII.I,ER Am(rag 1000 young adults who smoke cigarettes rc~darly, - at.xml 1 will be murdered - alxmt 6 will be killed on the roads - al~mt 250 will be killed by tobacco. "For example, one aright follow the main me,age with a few explanatory notes, such a~: - Some of th,-L,w kill~ hy I~bacco wmdd have di~l ~n anyway, hut other~ ndghl have lived m~ 5, 10, 20, 30, dr more extra yenr~; the average amount of life Io~t by them l~ing 10-15 year~. - If yo~ glvc np I~-fi~re yo~ have ~a ~i~nt~ hrml dilate, hmnchili~, or rnncer, Ihen yogi nlrr~l ,ff lhe ri~k of~'n~h from sm~king. - llzmnge Io I~ }~ly from smoking accurnnlale~, m linage w~ ~1~ in their l~u~ will I~ nt greale~l risk in mk~le a~. Even in the United Slates, where road accident death rates are more than double those in Britain mad murder rates are about ten times those in Britain, some such comparlso.ns may be helpful (ahhough it may then be advisable to start with only 100 young United Stales adults, and to threaten about I, 2, and 25 of them with death). Whatever format is preferred, however, the central point remains: The reason one wants to prevent smoking is not just because it is dangerous--dozens of things are dangerous--but because it is so dangerous. This indicates getting some sort of quantitative information over, both about the effects of smoking itself on mortality and, perhaps at least as importantly, about how much srnallcr all reliably known other carcinogenic effects are. Such information may in the short term make only a few people give up, but over a few years wide acceptance of such a pcrspcctlve may have substantial effects, either on individual behavior or on making other actions politically acceptable. Tar Levels The foregoing epidemiologic evidence (especially that on trends in lung cancer morlality among English men in early middle age) strongly suggests Ihat, even without any substantial changes in cigarette sales, practicable reductions in sales- weighted tar deliveries may well reduce the lung cancer mortality from smoking ~08~9~90~ The Conlrol of Lung Con~ er 17 sul~stanlially. There was, moreover, in Lee ant Garfinkel's review (20), no evidence that any other disease was aggravated b) such changes mnong smokers of similar numbers of cigarcttestL Also, although tar and nicotine decreases do prcxluce some compensation in tim manner in whi h cigarettes m:e smoked, they do not appear to be important determinants of wl -ther or not people smoke, or of the number of cigarettes that they smoke. Tar reductions can IJ¢ implemented with no ;ttbstant~al political problems (especially if they are done centrally, or at least ~ ithout advertising campaigns that may suggest misleadingly to non- or ex-smol ers that low-tar cigarettes are safe), for they do not adversely affect the groa, ers, manufacturers, taxers, distributors or advertisers, and the smokers apl car hardly to notice gradual changes in tar deliveries. It is, therefore, unfortunate that while the WttO and UIGC have organized several meetings on stunk ng avoidance, some of which have produced reports listing a variety of practice,/suggestions for govermnents or fur vohmtary groups In consider, no similar reports arc available to help governments accelerate tar reductions. "/'he prob ~'ms are, nf course, quite dif- ferent from one conntry to another, depending o t whether the country is a to- bacco grower, a cigarette manufacturer, an export i~r, or an importer, on whether cigarette mannfactt,rc or distribution is virtually ia govermncnt monopoly, on whclhcr advertising is altowcd, on the era'rent tar I ,vels, and so on. For countries such as Brilain, dilli'renlial laxaliot| (which has ~]rcatly been used once sue- ccssfully to cut off the highest tar levels) could be 'l~scd again to cut off the next highest levels, and restrictions could be imposed im the advertising of, for ex- ample, all brands delivering more than 10 mg of t~, r. For countries such am China and Russia, where cigarettes are manufactured at d distributed by the slate with little or no advertising, and where typical tar levels exceed the upper limit of what is currendy sold in Britain, large changes could be ~roduced at little cost and with great benefit to their people early in the next con ury~s. However, the practical problems of how to help governments decrease ta " levels (without inadvertently encouraging a belief that low-tar cigarettes are sah ) is a large question, almost as deserving of carefully thought out, practical repot ts as the problems of smoking avoidance are. Ultimately, of course, the aim is to produce cit'~umstances in which very few people choose to smoke, but in a world where cig;wctte sales are still increasing, rather than decreasing it is not wise to let the perfc~ I be the enemy of the imssiblc. "In view of the extent to which smokers of low-tar cigarettes r o "compensate," however, and of the uncertainly as to which smoke components chiefly affect heal ! disease and chronic ol~tructive h,ng di.'~ease, there shmdd be no implicit expectation that these dis 'ases will also he avoided, and indeed the large case-cnntrot study of Kaufman et al (18) soggests ap woxlmate equivalence of the effects on myocardial infarction of the different types of cigarette thai ~ re currently available. ~During the 1950s, men in Finland still smoked "Russlan-st le" cigarettes, and in 1960 male hmg cancer iucideurc rates in early middle age were ~imilar in Bus :ia and in Finland. By the early 1980~. Imwcver, lyph'al Finnish rignrclle tar dcliverie~ had dr, pped m only 10-15 m~., whih" typical Rus- sian cigarette ~ar dcllvcric~ were slill about 20-30 my.. Recc ,t Finnish male hmg cancer ~midcncc rate~ in early middle age have decreased by nearly half, whil' those in Russia have hardly ahcrvd.

18 f'dchard Polo and I'dchatd Doll Table 6. lnfi+rmallon for governments on simple measures fi~r the control ofhmg cancera Price increases will pnxltwe fewer deaths anti more revenue (as long as they do nt~t feed Imrk into wage demands). Tar reduction~ ~uld ~ enconra# (e~p~iMly in countrie~ ~och as Russia and China whe~ typical tar levels are still of the or~r of ~30 rag, which i~ ext~mely high). Advertifing could ~ t~ed, ~t~ct~, prohibited, or limit~ to ciga~ttes delivering un~r I0 mg tar. Simple, clear, quantitative information could t~ communicated effectively to the general ~pulatkm: ABOUT A QUARTER OF ALL REGULAR SMOKERS ARE KILLE1) BEFORE THEIR TIME BY TOBACCO. GeneraJ no e: rvt on menda mn of hose few stmple measures (whwh might have a mtb~ an ial cffec m just a few off ex~) thm~ nt~, of court, detract from the need for a wide range of other meamtre~, including many ofth~" integrated by W}IO (2). UICC (6), and Tar-level reductions are not the only simple possibility for governments (Table 6), and they may do little for vascular or respiratory disease. But they tnay well offer one of the more immediately practicable means of avoiding an appreciable proportion of the mass of lung cancer deaths that can otherwise be cxpcclcd In oc- cur dttring Ihe first few dccatles of the next century. References 1. Parkin DM, Stjernsward J, Muir CS. Estimates of cancer occurrence througbout the world. WHO Bulletin (in press). 2. World Health Organization. Gontrolling the smoking epidemic: report of the WHO.expert com- mittee on smoking control, tFechnical Report Series 636. Geneva: WHO, 1979. 3. U.S. Surgeon-General. Smoking and health. U.S. Department of Health, Education and Welfare Publ No (PHS) 79-50066. Washington, DC: U.S. Government Printing Office, 1979. 4. U.S. Surgeon-General. The Health Gonsequences of Smoking--Cancer. U.S. Department of Health Education and Welfare Publ No (PHS) 8'7-50179. Washington, DC: U.S. Government Printing Office, 1982. 5. Doll R, Prig R. Quantitative estimates of avoldable risks of cancer in the United States today. JNCI 1981.66:1191-1308. 6. Gray N, ed. Lung Cancer Prevention: Guidelines for smoking control. Geneva: Union lnterna- tionale Contre le Cancer, 1977. 7. Ontario Council of Health. Smoking and health in Ontario: a need for balance. Report of the Task Force on Smoking of the Ontario Council of Health. Toronto: Ontario Government Bookstore, 1982. B. Doll R, Peto R. Cigarette smoking and bronchial carcinoma: dose and time relationships among regular smokers and lifelong non-smokers. J Epidemiol Community Heahh 1978; 32:303-13. 9. Doll R. The age distribution of cancer: implications for models of carcinogenesis (with discus- sion).J R Statist Soc A 1971; 134:133-6. 10. National Academy of Sciences-National Research Council. Reduced tar and nlcotine clgarcttes: smoking behavior and health. Washington, El'G: National Academy Press, 1982. I 1. CederI,Bf R, D,.~II R, Fowler B, Frieberg L, Nelson N, Vouk V. Air pollution and cancer: risk assessment methodology and epidemiological evidence. Environ Health Perspect, 1978; 22:I-12. 12. l)nll R, I'elo R. Mortality in relation |n snmklng: 20 years' ol)servallm;s nn male British din:lots. Br MedJ, 1976; 2:1525-36. 13. Wald N, I~ll R, Gopeland G. Trends in tar, nicotine, and carbon monoxide ylehls of UK cigarettes manufactured since 1934. Br Med J 1981; 1:763-5. 6~0~9~90~ The Cooled of Lung Cancer 19 14. Wald N, hlle M, !~reham J, Bailey V, Van Vunakls H. Serum codnine levels in pipe smokers; evidence against nicotlne as a cause nf ct~mnary heart disea~, l.ancet 1981 ; ii:775-7. 15. (:ederlfif R, Friberg I,, ltrub<'e Z, I,orieh U. The relatlon~hlp of smoking anti t~me rovariahles to mmtality and cancer mnrlfidity. Sto, rkholm: Karollnska Institute, 1975. 16. Selikoff 1.1. (:onstraiuls in estimating occopational cancer mortality. In Polo R, St hneiderman MA, ed~, Q.uanlilication of occupational cancer. New York: Cold Spting Itarbor Publications, 1981. 17. International Agency for Research on Cancer. Cancer incidence in llve continents, Vol III. Geneva: World tlealth Organization, 1976. 18. Kaufman DW, Helmrich SP, Rosenborg L, Miettlnen OS, Shapiro S. Nicotine and carbon monoxide content of cigarette smoke and the risk ofmyncardlal infarction in young men. N Engl J Med 1983; 308:409-413. 19. Wald N, Idle M, Boreham J, Bailey A. Inhaling habits among smokers of different types of cigarette. Thorax 1980; 35:925-8. 20. Lee PN, Garfinkel L. Mortality and type of cigarette smoked. J Epklemiol Community Health 1981; 35:16-22. 21. Lee PN. Statistics of smoking in the United Kingdom. Research Paper No l, London: Tobacco Research Council, 1976. 2'2. Auerbach O, Hammond EC, Garfinkel L. Changes in bronchial epithelinm in relation to cigarette smoking 1955-19~J0 versus 1970-1977. N Engl J Med 1979; 300:381-6. 23. Doll R, Peto R. (1983). Epidemiology of cancer. In: Oxford Textbook of medicine. Weatherall I~I, l.edlngham J, Warrell I_)A, eds., Oxford: Oxford University Press, 1983: 4.51-4.79. 24+ l.lplml A. The bit~lngical effects of Io',v-level ionizing radiation. Sol Am 19'82; 246(2):29-37. 25. Garfinkel I,. Cam'er mortality in tmnsmokers: Prospeetlve Study by tile Amerk'an Cancer Fka'icty. J NCI 198t1; li5: I I lig- I 173.