Tobacco Documents Online

TAR CONTENT OF CIGA.RETT~$ N RELATION TO LUNG CANCER DAviD W. KAY.~ JULIE R. PALMEI~' LYI~ ROSE~rBERG,' PAUL STOLLEY.: ELLEN WAR~HAUEP~a ~ND SAMUEL Kmufman, D. W. (Sione Epidamiok~y Unit, Bostm~ U. ~1 of M~ic~e, B~kl~, ~ 02~), J. R. PaI~, ~ R~n~, P. S~ey, ~ Wamhmuer, ~d S. S~pim. Tar ~nt of ~s ~ ~on ~ lung ~. Am J E~ 198~129:70~11. ~ ~s d~s ~t y~ ~e ~ ~ fav~ ~ ~ h~s. ~e a~m eva~at~ ~Js J~ ~ a ~y o~ ~1 ~ who m ~ ~ ~ yearn ~ ~ ~ ~m ~ by ~w ~ ~s In ~ Un~ ~s a~ ~ ~ ~r 1~1 ~ June lg86. F~ ea~ ~ear ~ mo~, ~a~ b~ ~ ~nt as p~ in ~ ~ Tm~ 1985) ~ ~ ~s ~ (~ 1~ ~ 1~). Tar v~ m d~M~ a~ng ~ me ~r ~ of ~r ~ ~nds a~ over a s~ ~, in~ low (~ mg/~gam~), ~m (~-28 and h~h (~ mg/~ga~) ~ m~m, ~ ~ a~8 ~r ~t was es~s ~or m~Jum and ~ Mr ~am ~~ ~ ~s ~ ~r ~F ~ ~ was ~~ (p: 0,~2). a~ m~ and w~ ~ m~ ~de ~ ~ ~r ~ h~is ~ dam ~m ~m~ in ~st ~e ~m v~ ~w exam awmg~ less ~an 10 ~/~e. lung nemplesms; smoking; lars It i~ generally accepted that the risk of cigarettes, with smokers of low tar eiga- lung cancer is relat~ eo the tar content of ret~s having a higher risk than non- l:lae~ for px~l;ce~on Feb~'y 29. 1~, ~ ~om A~ 12, l~- Heath, ~n Oni~nity ~bool o£ M~e, ~ent Of Me~Cine, ~ion of M~ne, ~ ~olo~ Uni~ The of P~ylv~is ~l of M~e, P~e~ M~o~ SIo~-K~ C~r ~, N~ B~, ~ 0~1~, 708

smokers but a lower risk than smokers of high tar cigarettes. This has considerable biologic plausibility, because "tar" con- t~ns known can:ino~ens, ineludin~ benzo(~)pyrene, dil~uz(~)an~cene, ~- me~hylchr~sene, and o~her polyaroma~ic hydrocarbons (1). Fotrr epidemiol0gic s~ud- ias have been published with dam on ~ conten~ (2-5). L~ two, the results sugge_~taed that ~ is related to lunG cancer risk. The American Cancer Society conducted a large-scale folJow-up study in which over 1,000 cases of lung cancer developed during six years of follow-up (2). Among men, the standardized mortality ratio for smokers of low tar cigarettes compared with smokers of high tar cigarettes was 0.8; compared with smokers of medium mr cigarettes, i~ was 0.9. These results were not statistically significant. Among women, the correspond- Jug stanc~,~ mortadJty ratios were 0.6 and 0.7, respectively, and they were st~tis. tically significan£. In a case-control study of women from Austria, the relative risk estima~.s for low, medium, ~ud high tar smokers relative to nonsmokers we~ 2.6, 4.4, and ~.9, respectively (3). These esti- mates were based on 1B8 cases who were smokers, and the trend was sta~is£ic~y The ~sults of ~he other ~wo s-mc~ies were equivocal In the Whitehall s~'~tdy, after 10 years of follow-up there were 108 fatal c~ses of lung cancer among smokers who inhaled (4). For ~ose who smoked less t~han 20 prov~c]e~ the ps~ent~, t~ the nurse.interviewers ~ho ~ ~m ~a~m~c, and ~ Cyn~ia Whir for ~ foll~ h~ ~ci~ ~ the H~i~; New Yo~: Mem~ SI~-Ke~ C~- ~ H~, Thomas Jef~e~n U~ve~i~

No. ~ Neo % M~e 5~ 61 99~ 39 FemaJe 347 39 1~72 61 ~9 ~9 14 9~ ~ ~ 41 ~2 33 Non~i~ I~ 18 5~ Y~ of ~15 479 ~ ~ 52 ~16 I~7 17 714 Ci~ of h~pi~ N~ York ~7 ~ ~ 24 B~o~ ~ 7 ~ ~n 2 0 17 1 . ~n~n, Oa~o, C~a~ 1 0 7 0 and controls ~ccordh~ to v~iOUS factors. Compared with the controls, the cases were mo:e commonly male, ~omev,h~t older, and less we~l educated. Over 50 per cent of eases were interviewed in hospitals in Phi]- ,-delphla, end only 13 per cent in Boston or ]~altimore. By contrast, the controls ~ere more evenly d~:ributed geograp]~icaI]y. Definition of ~r exposure The information obtained on cigai~tte smoking included current smoking stat~s (smoked within the past year, |a~t smoked at least one year ago, or never smoked), number of c~ret't~ smoked per day, tom~ duration of smoking, name ~d number of years smoked of the current brand or the most recently smoked brand, and name of the brand smoked the longest. The bra~d names e~d durations were used together with taz values published annually by Federal Trade Cozn~iasion from 1967 to

i985 (there were no reports available for the years 1969, 1971. 1972, 1979. I9~0, and 1982) (9-18) and by the Reader's (based on similar lahorato~j, methods) for the yea~ 1957, 1958, 1959, 1961, 1963, and 1966 (19-~-4) t~ de~ermine an individual's tax e:k'):~s~e. Cigar a~d ]~pe smoking not considerecL It was first necessa~ to eat/mate values for the yea~s for which no report was avail- able; this was done by linear interpolation for the missing years after 1959. For the years before 1957, that is, before any ports were available, we maAe the assump- tion that the t~r values were the same as the earliest reported value for each ciga- ]~ette. (Lf anything, they were likely to have been higher, but there was no way to doc- ument this.) For each individual, we assigned a tar value for each yea~ of smoking. For some, the process was straightforward, because they only smoked one brand of cigarettes, or at most, two; that L% the sum of the durations of smoking the two' repo~lzd brands equalled the ~otal d-ration of smok- ing. When the sum of du~ations ~as tess then the total duration, we assumed that the brand smoked the longest was smoked in the middle of the period before the most recent brand was smokecL Tar values for th~ years between the assumed end smoking the longest-smoked brand and ~e beginning of smokgng the most brand were then. estimated b~ Linear inter- polation. Values for the period b~fore b~- ginning smoking of the long,st-smoked brand wer~ estimated to b~ equal to reported ~r values for tha~ bran& (This Last approach ~as also us~l Chen there only information on a single brand and the d~ca~io~ of smoking that bra~d did no~ equal the total duration.) Implicit in this p~cess was the fu~her assumption when subject~ s~it~hed brands, it ~as to a cigarett~ ~ith similar or lower ta~, and not to a higher tar bran& This assumption reasonabh on the basis of ou~ data~ Over SO per cent of the smokers who s~itched did so to a similar or lower ta~ brand.

707 spon~n~ ¢~ tar levels into the mult~vaziate m~iels. RESULTS A total of 614 of the ~I cases (70 per cent) and 85~ of the ~,570 conero|s (~ cant) were c~ent cigare~ ~moke~. As ~own in ruble 2, the es~a~ ~la~ve r~k ~c~as~ ~a~ly with ~c~g n~ber of cig~ smoked ~r ~y, ~m &0 for ~e Hgh~ amake~ (<15 c~s/~y) W 60 for ~e h~vie~ smoke~ (~45 te~/~y). Foz ex-smoke~, the w~ 6.8. For ~y current smo~ng, it 20. Smokers were &vided in~ t~ ca~- gozie~ according to averagz tar content of ci~are~s smoked, Relative risk estimates for the intermediate and highest ~ cate- gories compared with the lowest tar cate- gories are displayed in table 3 for subjects for whom the brand smoked was known for at lea~t 75 per cent of the toted duration of smoking, and for the subset for whom the brand smoked was known for the duration of smoking (100 per cent). ~n each in~tance, the estimated risk increased ~dth increasing tar conten~ although the esti- mates were somewhat lower for the 100 per cen~ subset: In the 75 per cent group, the z~imate for the in~ennecliate ta~ compszed with the lowest catego~ was 6~ 214 a_O 5J3-.]2, 20'/ 363 1S 10-.23 96 118 £8 17-44 149 17,0 43 27-.~ 9(; 4'/ CoO 35-102 7 ~4 6.1 ~.2-17 • l~fem.nce T~L¢ 3 Ju.n~ 1986 * l~eren~ cs~ory.

708 and for the highest tar category it ~as 81; the Wend was ~J~,nificm~t (X = 3.3, p 0.0~1). The corresponding estimates for the 100 per cent subset v, eze 1.5 and 2.4. re- spectively (X ffi 2~, p = 0.02). The remain- de£ of the results reported here am based on the enti~e 75 per cent population of smokers. Compared ~ith never smokers, the rela- tive risk estimates for smokers of the low- ,t, i~termediate, and highest t~r cigarettes ~ere II (95 per cent confidence interval (CI) &9-19), 23 (95 per cent CI 14-3S), and 36 (95 per cent CI 18-73), respectively. Med~n ~ values for the highest cate- gory were 311 rag~cigarette among cases and 31.6 rag/cigarette among controls, and the highest values were 34.0 and 34.1, spectively. The con~ndi~ me~e~s for the lowest categories were 19.1 and 18.6 mg/~ette, and the lo,vest values were 15.3 and 9.1, respectively. No cases and six contro|s (2 per cent of the smokers in ~he 75 per cent l~pulation) had average tar values of les~ than 10 rag/cigarette. As explained, average tar values were e/so computed separately for cigarettes smoked 10 or more years before hospital admission. The rest~s axe shown in 4. There were fewer cases and cont~o|s the lowest tar category and more in the highest tar category compared with the dis- tributions based on all years of smoking. The r~latlve risk estimates increased with increasing tat c~ntent, and were somewhat higher than the estimates based on all years of~moklnb-. For smokers in the intermedl- ate tar category, the estimate was 3.0 (cvm- pare~ ~ith I-9 from -ll yesr~ of smoking). and for tho~e in the high War category, it was 4.0 (compared with 3.1}. The u~nd was statistically significant (~: ~ 3.1, p--- 0.002). A f~xther analysis was ba~d on ciga- rettes smoked at lea-t 20 years before hos- pital adudssion. The results were not ma- terially different from those hasecl on ciga- rettes smoked at least 10 years previously, and the remaining analyses are based on the latter period. As is shoam in table 5, the pattern of increasing ~ with increasing tar level broadly consistent by sex. Among men. the relative risk estimates for the intermediate and high tar catsgorles ~ere 3.e and 4.0, respectively: the Wend was on the border- line of statistical significance (X = 2.0, p 0.05). Among ~omen, the corresponding estimates were 1.8 and 4.7; this ~rend was significant (~ -- 2.1, p -- 0.O4). The data for men and women were ther stratified by number of cigarettes smoked pe~ day. In most categories, the relative risk estimate tended to increase with increasing tar content However, num- bers were small and the result~ were statistically significant (data not shown). DISCUSSION The hypothesis that tar in cigare~ smoke is a major component in the etiology of lung cancer has biologic piausibil(ty and is supported by some epidemiolog~c data (2, 3. 6. 7), although the magnitude of the association has tended t~ be modest. In additivn, some results have been e~dvoca] (4, 5). The ~,lta of the present large control study provide further evidence that the War content of cigarette smoke is h~de~d T~Lz 4

~b. 18.1998' 0:~ N6~ 1153 P. 8/ZU I,~G CANC~ A~lD ¢IGM¢EI"~ TAR ~'09 ~v. car content Mu)r.1"vm'i~ mlaf.i~ M~ <22 ? 32 1.0" 22-28 68 48 3.6 1-2-11 relate| to ku~.g cancer risk: We estimated that current smokers who had smoked aretCes 10 or more years ~t the past ~dth an average of at least 29 mg of tar ]~r ciga~tte had approximately four ti~es the zlsk of lung cancer as those who had smokec] cigarott~ with an average tar con- tent of less than 22 rag, Smokers of ¢iga- Rcms with intermediate tar content ap- peared to have a threefold increase in r~k. The association was present in both men and women and c~d not appear to he counted for by number of cigarettes smoked per day. I~ interpreting the present ~-sults, considered the possibility of d~storcion by various forms of bias. Selection bias seems m~likely: Lun~ cancer is an illness for which diagnosis and admission to hospital is al- most inevitable and is certainly unrelated to the tar content o~ cigarettes smoked, especially in the distant past~ The controls wer~ selected to have diagnoses to smoIRng, some of which involved acute diseases requiring hospital admission. It is h~ghly unlikely that among smokers these conclitions would he ~elated to tax content. It is conceivable that controls tended to underreport high tar smoking, or cases low tar smoking, pmclucing a spurious associa- tion, b~t this does not seem likely. Poten- tially confounding factors wexe t~ken ~nte account in est£mating the ~elative risks, and Lncluded, in particular, the number of cig- a~ettes smoked per d~y and the duration ot smoking, Given the relatively uncompli- cated e~iology of lung ca~ce~, it does not appear Rkely that there was uncontrolled confounding by f~ctors not taken into ac- cotmt~ A weztkness in our study was that we did mot have complete lifetime histories o~ brands smoked for most subjects (siz~ee only the most recent and longest-smoked brand~ were obtained), nor were tar values available £or all years. The assumptions t2mt were made about the gaps in brand history, however, i~ incorrect, would have tended to underestimate average tar e:~po- suxe, which would have |ed to u~dsresti- marion of the relative risks. To zainimize the effect of such assu~nptlons, ~e confined the analysis to cases and controls £or whom there was information covering at ]east 75 per cent of the duration of smoking; that is, tar values were asslz~necl ~oz not more than 25 per cent of the smoking history. This did result in the exclusion of s sub- scantial proportion of subjects ~or whom there were larger gaps in the brand histo- ries, but there is no ~eason to think that the population an~lyzecI was biased with respect to tar val~es. We also analyzed the ~-ubset of cases and controls for whom there was brand iz~form~t~on on the entire dr|ra- tion of ~moking, and the results were sim- ~ to those for the 75 per cent population, although numbers (and the proportion of

dural{on of smoking, there ~ere very few subjecSs who~e lifetime tar exposure aver- ~ le~ th~n 10 mg/c{~m'et~e. Given the heat mar{mt{u~ of such cigarettes ~i~h the hnplication that they ~e safer and ~ven their widespread use, 9zture studies should be focused on this cap/n our knowledge. 1. Smokins aud h~Irh: a ~x~re of ~ S~n Gen- t. W~ DC: ~p~ant of H~h, ~o~ ~d W¢~, 1979. ~ pub~on no. (P~)~-~). ~d~ W d~th ra~s. ~v~n ~ 1976;12:2~ 3. V~ C, K~ M. ~ ~ ~on m ~ ~1~ of I~II:71~18. ]~ ~, ~d ~ h~ ~: ~e e~- f~ of ~a~on ~d ~ ~el~ ~ Epi~0I o~r ~ ~ ~oke~ o~ low ~eld ~. of 1~ ~ ~ ~t~ ~ok~ Envi~n fie~ a ~n~ol study. JNC] 19~2:471-7. • Slon~ D, Sh~ S. ~e. OS. ~ ~l~ ~d ~e~ A~ W~n, ~: F~ T~ ~{on. 1967, 1~, 19~0, 1971, 1973. F~r~ ~a& ~#sio.. 1974. 11. ~ of~" ~d ~c~e w I~ v~e~ ~ c~re~. W~n, DC: F~eral Tmd~ ~m~ion, 19q5. 12. ~ of~r" ~d nicene ~n~n~ of~e smoke o~ 145 va~e1[~s of c~a~z. W~in~n, DC: F~r~ Trade ~ion, 13. ~ of ~r" ~d ~ne ~n~nz of the ~oke 15. ~on of ~ ~," ~t~e, ~d c~= {~ ~n~n~ of ~ smoke of ~ W~n, DO; F~eral ~a~ ~mmi~- ~o~ 1981.

711