Tobacco Documents Online

PI~VENTIVE ~4EDtCIN~ 17, 26~.-.272 (19~) NO. 1153 p. 11/20 Smoking and Lung Cancer: Risk as a Function of Cigarette Tar Content~ HOHEt~ B. WILCOX, M,S.."-" JANET B. SCHOENBERG, THOH.~,S ~. ~.*SON. PH.D..+'z ]oA~s ~ ~ ANNE~E STEHHAGEN, Da.P.H.*'~ "Cancer E~idemiolo~y ~emice~. DiviMon of Epidemiolo~y Department of Heal~k, C~ 369, Teemon. New Jersey ~62~. and tE~yi~onmemtal Branch. Divixlon of Cancer EHology. National Cancer The hypothesis of reduction in lung cancer risk associated with the adoption of low-tar cigarettes was =xamined in a subset of subjects from a population-based, case-control study or in¢idtm primary lung cancer among New Jersey white men. Risk was related ~o time- weighted average tar levels of cigarettes smoked in 1973-1980. Unadjusted estimates of risk were significantly low for the Iowesz tar (< 14 mg/ci$} smokers [odds ratio = 0.53 (0.29,0.97)] compared with the highest (21.1-28 mg/¢ig). However. adjustment by age and total pack- years rendered the risk reduction insignificant. Of note was the finding that cases who smoked low-tar cigarettes compensated for reducing tar by increasing the number of ciga- rettes they smoked by almost half a pack per day from the years 1963-1972 to 1973-1980, while in the same period controls and high-~a~ cigarette smoking cases did not inorease the numbers smoked. © ~ ~me~= pr~a, |~. INTRODUCTION Cigarette smoking is generally recognized as the primary cause of lung cancer (23-25). Tar, the total dry particulate componem of smoke, is a complex mixture of hundreds of substances. There is a body of information classifying these mr corapounds as initiators, promoters, cocarcinogens, and so on, and increased consensus as to how these raolecules act (28, 31). As to the general, conclusion that cigarette tar constituents increase cancer risk, there is no dispute (25). The quantity of tar per average cigarette market¢d in ~e United States has be¢n declining steadily over three decades for two reasons. The mix of product pur- chased by the smoking public has shifted, first as filter cigarettes became preva- lent in the 1950s, and subsequently as companies introduced new brands to corn- pete for the perceived "'low-tar" market, Also, tar reductions have occurred as manufacturers reformulated cigarette brands continuously available frora before 1950 to the present. Before 19.53, the average tar content of cigarettes was ap- proximately,35 ing each. By 1962, tar in nonfdter cigarettes had declined to about 30 mg each and filters to 22 rag. After another decade, non-filters tested at 27 rag and filters at 17-18 nag. In 1980, non-fthtrs averaged 24 rag and filters ordy 11 rag, This study was funded by Contract Noi-CPdil031 from the Nadona~ Cancer Institute. To whom reprint requests should be addressed. Current address: Fox Chase Cancer Center, Philad¢lph/a, PA 19111. Current address: University of Pennsylvania, Ph/ladelphia, PA 19104. 263

264" WILCOX ET AL. - .... Tl~re£or~, one ~[ hypothes~e some co~es~udins, m~u~ble impa~z o~ this d~in8 ~ coment on ]un~ ~c~r iucid~ or ~sk. ~s m~y ~ ph~d o~e o~ ~o ways: (a) is there de~r~ed lu~ ~cer ~sk ~or ~erso~s who smoke low-~ cig~ect~s exclusively; ~d (6) is present smo~ng of lower-tar cig~ttes, ~tcr some te~ of smo~ng hi~er-~ ones, ~s~iated wi~ decreased lung c~- cer ~sk? We have inves~gated the latter using da~ ~om a ~pulation-based ~cident l~g c~cer c~e~on~rol inte~ie~ study of w~ce men ~om six ~eas of New ~ersey. M~HODS ~ ~cidem white m~e [uag ¢~acer c~es ~osed .d~g the pefi~ o£ eem~r 1980 to Octo~r ~98], ~d ~¢~ ~n s~ pmviously aoted hi,-mothy ~ of Hew Jersey (2~). were iden~. ~es were ~c¢~¢d vi~ s ~pid ~~ sys[cm es~lJshed ~tw~e ~e ~ew 3e~ey ~p~ment of ~¢~ J~ hospJ~ pathology dep~mems, by few ~view of hospi~ pathology lo=s, ~d by mo~to~ ~e S~te ~r ~¢~s~ =d d~ ce~cste Fdes. O~y ~sto]o~y co~=d Qses w~ re~ for study. Popu~on-b~ed ~atro]s were select~ ~ ~o pro~dures, Coa~ols £or dk~tly ~ewiew~ c~¢s were sei~t~ ~m • ~dom s~pl¢ of m¢~ ~th New 3ersey motor vehicle fic¢nses. ~s s~ple w~ fr~ue~cy m~tched m ~es ~eo~p~c ~, ~e, ~d 5-ye~ ~e ~oup. Con~]s for de~ or ~ ~s were sel¢ct~ from s=¢¢ death ce~¢ ~¢s, ~d we~ ia~vidu~y ~tch~ m c~es by ~a, ~, ~¢, ~d closest dste of d~ (or ~te of de~ ~st ~¢ ~¢ of di~osis, for in~i~t~ c~¢s). Ce~tes ~c~g ~u~ c~r or ~y other resp~¢ow d~se~sc ~ ~ uaderly~= or coa~bu¢ow cause o~ d~ were excluded. ~t~¢ ] ,~ whit¢ m~¢ lung c~cer c~scs efi~bl¢ for ~h¢ study, i~z¢wiews were success~gy completed ~or 76~ (70.4~), 429 ~th se~-res~ndents ~d ~4 were la¢c ~ce~nmcnt (6.9%), ~su~cient dine ~¢r d~th to ~]ow the pe~ reread by ~he Insti~ution~ ~ev=cw ~oard before contacting (2.1%), physician refusal (4.2%), un¢~ce~bJe ~spon~¢~ (4.2%), Rspondcnt ~us~ (]0.B%), and pooPqu~ity {nterviews (~.4%). ~ the ],415 whj~¢ m~e con- trois ~dent~¢d, interviews were successfully completed w~th ~ (63.6%). wi~ se~-respo~dcms a~d ~36 whh next-o~-kJn (3 ]8 first- de~ec rciazi yes, ~ ~ more distain). ~e~so~s ~or nonincl~sio~ w=~¢ un~bl¢ ~spon~em (9.0%), respon- deal r¢~s~ (24.7%), =nd poor-qu~ky ~¢wiews (2.8%). ~ Jntewiews o~ subjects or thek next~f-~n were conducted in person by s~ expe~¢nc¢d in larg¢-~c~¢ epidcmiolo~c s¢udics. ~¢ modular included sections on demographic ~actors, residential h~s~ory, personal and J~l mcdicai histow, histow of cisare~te, pipe. and ciB~ smoking, smoking ~y other household members, occupational hJs¢ow (20). ~nd dietary history In the cJgare[t¢ use section o~ [he questionnaire, a smoker ~rst wa~ ~c~ recall [he ye~s in which h¢ smoked c~g~r¢~tes ~or ~ny pc~od ~ months or

,Feb.

No. I153 P. ~,b. tS.[£98 10'21A~ 266 WILCOX ET AL .... • ;.. ' TABLE I CHMMCT~mSTmS OF CASE mW~) CoN-r~oL CiG^~'rr~ SNOKF.~S, Wt~t Succ~ssrvF. Rt~s-rmc'rmNs OF SAMPLE, NEw JEF~SEY WHIT~ MEN, 19B0-1981" Subset A: a/i Subset B: all in All cigtrctte with complete A wBo smoked smokers smok~n| histories r~roughout 19"/~--19S0 Cases Controls Czses Controls Cases Controls N umber of ~u~jects 727 671 583 559 373 247 Age, years 64.7 64.4 64.2 6~. 1 62.1 62.5 Intensity, cigarettes/day 29.3 24,7" 29.4 23.8" 29.8 23.8* Duration. years ~3.5 36.4* 42.6 35.7* 46jr 45.0 Smoking cessation, years 4.7 10.1" 5,0 10,5" 0.043 0.061 Age bega~t smoking, years 16.5 17.6" 16.6 17.6" 1~.7 I6.9" Total expostare, packycars 64.4 46.2" 63.0 43.7" 68.8 53.2" Mean value. , P < 0.05 for difference between means of ~ascs and coetruls. sample and in the first subset; there were no caso-control age differences, Further restricting the sample to those who smoked throughout the inter~al 1973-1980 (Table 1, subset B) reduced the mean number of year~ since cessation to nearly zero, and erased the previously noted caso-control difference for duration. The remaining analyses refer only to these 620 subjects. Due to the reduction in tar content of both non-Filter and t'dter ci$arettes, the average tar level of the cigarettes smoked during 1973-1980 did not exceed 28 rag, 80% of the pre-1933 level of 35 mg tar per cigarette. The median tar level for the 373 cases was 18.6 mg/cig and for the 247 controls was 18.2 mg/cig. Four tar-level groups were selected. The frequency distribution of tar levels for controls showed three peaks, one consisting of the subjects smoking cigarettes with 21-25 mg tar, another consisting of the very-low-tar smokers, at 14 mg or less, and the bulk of the smokers in a broad middle group that we divided just above the modal value of 17.5 mg/cig. Table 2 shows that the unadjusted odds ratio for subjects at the lowest tar level, relative to subjects at the highest level, was significandy lower than 1.0. There was a significant trend (one-sided P = 0.04) in odds ratios with increasing tar. However, taking cognizance of the sub- ject's lifetime smoking behavior increased the odds ratio for the lowest-tar group so that it was no longer significantly different from the reference level. Controlling for packyears resulted in the greatest shift of the odds ratio toward 1-0. We examined possible differences in the packyear-~justed tar-level odds ratios by respondent type. Although self-respondents showed a more consistent pattern in odds ratios with increasing tar, the trend was not significant; none of the odds ratios was significantly different from the reference level, and there was no sig- nificant heterogeneity by respondent type at any tar level. The packyear and respondent type-adjusted odds ratios were almost identical to those previously shown in Table 2 for adjustment by packyears alone. We made similar comparisons of age-specific odds ratios, because the associ- ation with tar might be more pronounced in younger persons, with a shorter early

SMO~NG AND LUNG CANCER 267 TABLE 2 "" ' ~" " ODDS ~;~,AT|O~ FOR LUnG C~CEA A~IA~D ~ T~A LE~L OF C[o~ SMO~D [~ .Number o( ca~c~icon~,rols: L nadlu~ted odd= r.~tlo 0 .<3 0.B6 1.04 |.00 Adjusted by ~mnsJty~ 0.61 1.01 I. 16 (0.33,1.12) (0.~,1.~I) (0.72.1.~) Adjusl~ by du~tion~ 0.~8 0.89 1.05 1.~ {0.~2. ).o7) [o.~.l.32) (o.~. J.67) A~us~ed by inten~hy 0.61 I.~ 1.21 1.~ and dumdo~ ~0.32, I. l ~} (0.70. I .~6} (0.7~, ~ .~) Adjusted by ~ky~ars~ 0,71 l.~ I. 19 l.~ (0,37.1,35) (0.71.~.59) (0.73.1.93) • Sample inclUdeS smok©rs with complete smoldng his¢orie~ w~ smok~ ~mu~ouc 197~i9~ (subsm E in T~I= I). ~ ~m~Ha¢~¢l estate of ~ ~o (95% co.donee ~). ~ R=f=~¢ ~up. a As <20 ¢iW~y, ~29, ~]9, ~9. 50+, ~ ~ensiw ~ for ~, d~Uon <~ y~, ~ +. z As <~ ~ky~, 2~39, ~59. ~, ~. 1~ +. history of smoking the ava/lable h/gh-rar cigarettes. There was I/tile consistent vaz/a6on by age; the youngest age quart/le showed no paz~em in risk with increas- ing tar. The packyear and age-adjusted odds r~tios were affa/n almost identical to those shown in Table 2, except that the odds ratio for the lowcst-t~r group in- creased further to 0.$5 (0.42, 1.75). Because vegetable consumption is associated with lung cancer most strongly in current and recent ex-smokcn Oo), we were concerned ~hat the effect of diet might mask an association with tar, Therefore we calc~ated packyear-adjusted odds ratios by tar level and v©gcrable consumption, relative to subjects in r.he ll/ghest-tar/lowest vegetable consumption group. The associar.ion with tar was apparent only in the group with highest vegetable consumption, among whom the lowest-tar subgroup showed a sign/ficandy low odds ratio of 0.24 (0.07, 0.83). The trend in odds ratios wi~h increasing tar in ~his high veger~ble group had a one- sided P value of 0,~. The resul! of adjustmen[ by packyears and by vegD~ble consumption shnul~neously was no different ~an obta/ned from adjustmem by packyears alone; there was little consistent overall association with tar. Further- more, adjusul~ent by r~r level and packycars did not change the significant invers¢ association of lung cancer with vegetable consumption (32). Adjustmen[ by o~er pmen~inJ confounders, for example degree of inhalation or education in addition to pa~kyears, did not appreciably change any of the patterns in odds ratios previously descr/bed. The same was true of ~he |ogistic regression analysis. Examination of different combinations of factors suggested that signif-

18. lgg8 IO:21A~ No. 1153 P. IB/20 268 wILCOX ET leant contributors to the predictive power of the model were packyears, age, vegetable consumption, and employment in high-risk occupations. Inclusion of these resulted in odds ratios of 0.83 (0.43, 1 £0), I. 11 (0.74, 1.67), 1,28 (0.78, 2.10), and 1, for the four tar levels. It is possible that smokers, although receiving some unspecified amount of risk reduction by using lower-tar cigarettes, could negate that benefit by smoking more cigarettes. We compared, for cas¢s and controls at each tar level, differences between the numbers of cigarettes smoked per day in the interval 1973-I980 and the intensity in the immediately preceding decade (Table 3). Overall. cases in- creased their consumption more than controls did, although this difference was not significant. Among cases, there was a consistent increase in intensity with decreasing tar, whereas among controls there was no consistent pattern. The difference between case and control changes in consumption was significant for the lowest-tar group. We further attempted to localize this difference to a panic- ular tar level of cigarettes smoked in 1963-1972. The only significant difference in changes in numbers of cigarettes per clay between cases and controls was that for smokers who were in the lowest-r~r category in both 1963-1972 and 1973--1980. These 19 cases increased their consumption by 9.3 cigarettes per day between ~c two time periods, whereas the t3 controls showed no change whatever in con- sumption. DISCUSSION In addressing the question of whether smoking of low-tar cigarettes is associ- ated with decreased lung cancer risk, the cardinal problem is one of definition'. what is a low-tar cigarette? The way this problem has been approached by many is to differentiate cigarettes by a proxy variable, whether they are filtered or not (2, 12, 29, 30), While the amount of tar in filter cigarettes as a class probably has never been greater than chat in nonfdter cigarettes, it should be noted that filter cigarettes initially available in the early 1950s had higher tar content than plain tip ones today; "'high" and "low" are relative terms. An alternative approach used by others (8, I0, 13, 26), and with which wc TABLE 3 I~I~2, ~v C~E~o~oL STA~S, A~V ~v ~Aa L~V~L i~ ~97~198~'# T~r level, 1973-1980. rag/cigarette ~ 14.0 14. i-17,5 17.6-21.0 21.1-28.0 AB Cases +S.$" (25) +2.I (130) +0.7 (78) 0.0 039) + 1.4 (372) Cot~trols -l.l ('27) +0.8 (~9) 4. 1.3 (4~) -0.] (83) ÷0.4 {244) " Difference ,== No. elks/days (1973-1980) - No. cigs/day (1963-1972). No, of subjects in paren- theses. ~ Total No. of subjects = 616; 4 subjects who smoked throughout 1973-1980 did not smoke through- out 1963--1972. " P < 0.05 for difference in mean difference between cases and controls,

269 associate ourselves in this study, is to derive a scale from cigarette tea'-content information. Although some controversy surrounds the Federal Trade Commis- sion's methods and publ/shed lists of smoke component measures, the ordering of particular brands with respect to others is likely to be correct, and per extenso the allocation of smokers into average tar level groups. There are only a few studies that address the question of lou~er risk from lower tar by investigating ~xclusive users of either low- or high-tar cigarettes (more usually, lifelong filter vs plain tip). In the age groups of high lung cancer incidence. it is not easy to accumulate numbers of smokers of "'pure" type, who have never changed from plain to f~ter tip, Vmuc and Kunze (26) reported a risk reduction of 59% for their ndddl¢ tar-contem ca~egor? (corresponding roughly to our two middle tar-l,-vel groups, combined), compar~i with their high one (corresponding to our highest tar-level group}, with adjustment for age, dur'4tion, and intensity of smoking. The authors have included former smokers in their sample without adjusting for cessation: this complicates the issue since if smokers stopp,-d a numl~r of years ago, then they could not have b~¢n smoking the lower-tar ciga- rettes; on the other ha.ud, even short periods of cessation may far outweigh the effect of tar reduction. The parent data set of which the Vutu~ and Kunze sample comprises a subset has been mor~ extensively analyzed by Lubin and co-workers (12). Much of this work addressed l'dter/nortfilter di;[ferences, with attendant definitional di~culties. Restricting themselves to smokers never quitting, they found the equivalent of a 56% reduction in risk for those ~ho smoked only filter cigarettes compared with those who smoked only non-filters, adjusted for duration of smoking; they stated that there was little change after adjustment for age or intensity. While studies such as the t,~o above offer a sense of methodological purity in their effort to compar~ undiluted types (lifelong "'low-tar," "'filter,'" "high-tar," or "'plain"), they ar~ difficuh to interpret because populations of lifetime filter- only (low-car) smokers differ sharply from pla/n.only (high-tar) smokers. Specif- ically, among our control smokers (subset B, Table I), the I I Lifelong filter smok- ers were significantly different from the 77 lifelong non-filter smokers in the re- 'spect that they averaged 10 years younger, started smoking 5 years later, smoked almost a half a pack less per day, ~nd overall had one-third th~ total packyears exposure. These results are similar to those found in a large series of smoking histories recently assemb|ed by Wynder el el. (27), Clearly, behaviora~ aspects of smoking a~d age of the smoker need to be controlled car~fully, especiatly in any comparison oi~ users of exclusively one type of cigarette with another ~ype, The major/ty of studies on lung cancer risk in relation to tar content have, as in our study, attempted to quantify an hypothesized decreased risk among persons presently smoking low-tar (or filter) cigarettes, after some term of smoking higher z~" (or nonfilter) cigarettes. Some of this work suggested a substantial r/sk reduc- tion. gross and Gibson (2) exaa'nined filter vs non-t-flter cigarette users as by the type currently used. Filter users showed a risk reduction of 4~% with adjustment for both duration and intensity of smoking; age was not controlled in the analysis, although the entire: study saanple of controls was age-matched

No. 1153 p. IB/20 270 ~'ILCOX [~T AL ..... ;. : cas¢s. Wynder er aL (28), who likewise age-matched and adjusted for intensity, found a risk reduction of 41% for their subjects who smoked f'flters for at least 10 years, relative to plain-tip smokers. Other studies have suggested a smaller reduction in risk. Wynder and Stellman (30) again compared long-term f'dter smokers with non-f'dter users, but adjusted simultaneously for intensity and either age or duration; the reduction for the former group was either 16 or 21%, respectively. Likewise, Hammond er al.'s prospective study (8), matching subjects on an extensive variable list, yielded almost identical figures for risk reduction of lung cancer among smokers of low-tar vs high-tar cigarettes, 17% for the 1960-I966 part of the study, 21¢~ for the 1966- 1972 part. Lubin er el. compared long-term low-tar brand users with long-term high-tar brand users, and found a risk reduction of about 30% (I3). This analysis was controlled for number of cigarettes, length of habit, and cessation; the cases and controls being originally age-matched, apparently no adjustment for age was done. In our study, examination of smoking histories and the distribution of average tar levels convinc.ed us that the defmed groups reflect real classes of behavior, especially in 1973-1980: persons in the highest reference group still smoked non- filter cigarettes; the majority, in the two middle groups of the distribution, were smokers of popular and long-established (but changing over time) brands of filter cigarettes; those in the lowest group changed to the lowest-tar cigarettes avail- able. The initial unadjusted result of Table 2 indicated a risk reduction of half for persons smoking a cigarette in the 14 mg or lower tar range. However, any adjustment for lifetime smoking rendered the apparent risk reduction nonsignifi- cant. Inclusion of other confounders such as age decreased the risk reduction further, to only I7% for the lowest-tar group, a point at which it can hardly even be characterized as nonsignificant but suggestive. As Doll and Peto (3) have concluded, smokers in the age groups that contribute the great bulk of lung cancer cases appear to have been in fact so early and so intensely exposed to tobacco smoke that the only form of tar reduction that g~ves any significant benefit to them is complete cessation. Kunze and Vumc (I0) arrrived at much the same conclu- sion. One simple explanation for our observed absence of significant risk reduction could be that the theoretical benefit of reduction of intake of proven harmful material per cigarette smoked is offset by smoking more cigarettes. Other inves- tigators have surmised variously, for smokers switching to lower-tar cigarettes, that there is no compensatory increase (6, 29), that there might be a small increase (9, I 1), that there certainly is an increase (17,'27), or the their data are inappro- priate for the question (26); the Surgeon General urges further research (2a). Our data indicate that a subgroup of smokers in this study distinctly alters its behavior in such a way as to diminish the putative benefit: cases who smoked cigarettes with the lowest available tar for nearly 20 years also increased the number they smoked by almost half a pack per day. However, at the risk of advancing a post hoc propter hoc argument, the fact that controls did not show similar compensa- tion suggests that tar reduction could still be of some benefit.

'" 27I In recent years considerable discussion has been devoted to the proposition that, since smokers ~e not going to quit in large numbers no matter how grim the potential health consequences, perhaps some benefit might be gained by promot- ing b~nd-switching, or by redesign/ng cigarettes (7, 24). The implication of our finding of demonstrated significant compensation is that public health efforts aimed at converting smokers to the use of low-tar cigarettes must simultaneously emphasize the need to avoid smoking more cigarettes. Moreover. should there still b~e ~y doubt, efforts to promote smoking cessation and to prevent smoking zmtla~on should be emphasz.ze.d and intensified however possible. ACKNOWLEDGMENTS acknowledge he support of Dr. Ron~dd Alunan {New Jersey Sta~e Depa.mnem of Hc~ffi), ~e c~fion of ~1 New Jemey ~olo~sts in ~pid c~e ~¢e~nmen[. ~d the =x~nis= of ~he ime~iewcrs from ~c Cancer Epidemiology ~o~m. N/SDH. We ~ank ~, ~nda W. PieCe ~d Re~na Zither (Enviro~en~ Epid~ology B~ch, N~tion~ C~ccr ~sdtutc) for pmvi~ng ~c inspi~ion for ~e me~oio~ ~ ~e~ ~yses ~d for development ~ ~e ~e~ ~o~ation. ~s~cdvely. REFERENCES I. Bresiow. N. E., and Day, N. I:.. "Statistical Methods/n C~cer Rese~h: ~e A~ys~s of~ con~i Stu~es," Vol. I. ~tc~don~ A~ncy for Rash on C~r. Lyon. 19~. 2. Bross, L D. ~., ~d Gib~n. R. ~sk of lu~ c~c~ in smoken who s~tch to ~flter Amer. J. Public Health 58, 13~I~ (!~). 3. ~H, R., ~d P¢~. R. "'~e Ca~es of ~r: Q~dmdve Estimates of Avoi~le ~sks C~¢er ~ ~e U~t~ S~es T~y." O~o~ U~v, ~ss, New Yo~, 4. F~ T~e ~ssion. '" 'T~' ~ Ni~tine ~d C~ Monoxi~ of ~e Sm~e ~ V~eti~s ~ ~mesdc Ci~ttei." U.~. Gove~ent ~nd~ ~ce, W~n~on, ~, I~3. 5, F~e~ T~de Com~sion "Rein of "T~' ~d Ni~e C~t~nt of ~e Smoke of 169 V~etics of C~ttes.'" U.S. Gcve~ent ~g ~ce, W~on, ~. 1976. 6. ~¢1, L. ~ges in humor of ci~nes s~kcd ~mp~ m =~ges in ~ ~ ~cot~e co,tent o~r a 13-ye~ ~, [n "'B~ ~e~ No. 3; A S~e Ci~tee?'" {B. G. God ~d F. G. B~k. Eds.), pp. 1~28. Cold Spd~ H~r ~¢ow, Cold Spine H~r, NY, 7, Go~. B. G., ~d B~k, F. G. (~.) "'~b~ ~e~n No. 3: A S~e ~¢?" Cold H~r ~mW, Co~d S~ H~r. ~, 19~. 8. H~mond. E. C., G~nk¢l, L., Sei~, H.. ~d ~w, E. A. Some ~m ~n~gs ci~ne smo~ng, i~ "O~ns o~ Hu~ C~r,'" Cold Spd~ H~r Co~e~n~ nn Cell ~o[~e~ion, VoL 4, pp. IO1-112. Cold Sp~ng H~r ~bo~ow, Cold Spring H~or, NY. 9. Hoffm~n. D,, T~, T. C., ~d Go~, G. ~, ~¢ less h~l ci~[e. ~re~. Med. 9, 287-2~ lO. Kunze. M.. ~d Vueu¢, C. ~shoid of ~ exist: ~ys~ of smo~n~ ~smW of m~e lung ¢~cer ~es ~ ~n~b. in "'B~buw Rein No. 3: A S~e Ci~e?'" (B. G. ~ ~d F. G. B~k. ~s.), pp. 2~36. Cold S~ H~r ~mto~. Cold Sp~ H~r, NY, ~ L Lee. P. N., ~d G~nkei, L. Mo~y ~d ¢y~ of ci~¢[e smoked. I. Ep[dem~oL Comm. Health 35, 1~ (l~l). 12. Lubin, ~. H., Blot, w. 3., Be~o, F., ~t, R,, Gifl~. C. R., Kunze. M.. Schmahl. D.. ~d Visco. O. M~[fying ~he ~sk of develoVing lung c~er ~y chanong habits of ci~recte smoking. Brit. Med. J. ~. 195~1956 { I~). 13. Lubm. J. H.. BIoL W. J.. Be~no. F.. Fl~t. R.. ~illi5. C. R,, Knee. M,. Schmahl. D., ~nd Vis¢o. G. Pacte~s of lung cancer ~sk ~co~ing to cy~ of cigarette smoked, l~t. ~. Ca~c~ 56~576 { 19~).

No. 1153 p. ~IZO ," F~:b. 18. 1998 I0:23A~ 272 WILCOX I~T AL ..... .~. : 14. Mantel. N. Chi-squaxe test with one degrc© of'freedom: Extensioo of the Mantel-Henszel proce- dure. J. Amcr. 5tat. A~soc. 59. 690-700 (1963). 15. Mantel. N.. and Haensz~l, W. Statistical aspects of the m~alysis of data from retrospective studies of disease. J. Nat[, Cancer ]nst, 22, 719-.7~18 (1959}, lfi. Miettinen. O. Estimabifity and estimation in case-referent studies. Amer. J. Epidcmiol. I03, 226.- 235 (1976). 17. Petitti, D. B.. and Fr~edman. G. D. Evidence for ¢om~nsation in smokers of low yield Int. J. Epidemiol. 12, ~g7.-~89 18. SAS Institute'. "$A$ Supplemental Library User's Guide," SAS Institute, C~ry, NC, 1980. 19. SAS Institute. "-SAS U-~er's Guide: Statistics.'" SAS Institute, Cary. NC, 1992. 20. $¢hoenberg, J., Stemhagen, A., Mason, T. I, Bill,.I., mad Airman, R. Occupation and lung cancer risk among blew Jersey white males. J. Natl. Cancer l~tt. 79, 13.21 (1997). 2l. Stemhagen. A., Mogielnicki, ^. P., AlUnan, R., and Mason, T. J. "'Descriptive Fpidemiolog, y of Cancer Morr~lhy in New .lerscy. Vol. 1: Discussion and Analysis." New Jersey State Depart- ment of Health. Trenton, N.I, 19111. 22. Tobac¢o Merchants Association of the U.$. The TMA directory of cigarette brand names 1913. 1977. New York, 1975. 23. U.S. Department of Health and Human Services. "The Health Consequences of Smoking~ Cancer. A Report of the Surgeon General." [DH/:IS(PHS)g2501"/9], U.$. Govt. Printing Office Washington, D.C., 1982. 24. U.$. Department of Health and Human Services. "The H~alth Consequences of Smoking--The Changing Cig:~rette. A Report of the Surgeon General." [DHH$(PHS)gi-50156], U.$. Govt. Printing Ol'fice Washington, D.C., 1981. 2.5. U.S. Department of Health. Education, and Welfare. "'Smoking and Health. A Report of the Surgeon General." [DHEW(PH$)79-~0066]. U.S. Go~t. Printing Office Washington D.C., 1979. 25. Vutu¢, C., and Kun~, M. Ta~ yields of clgarette~ and male lung ¢.ancer risk. J. Natl. Cancerlnst. 7I, 43.$.-~,37 (19~3). 27. Wyndnr, E. L., Goodman, M. T., and Hoffmann, D. Demograpkic aspects of the low-Yield cig- arette: Cons/derations in the evaluation of health risk. J. Natl. Cancer lest. "/2, 817-822 (19g4). 2~. Wynder, E. L., nnd Hoffmann, D. Tobacco and tobit.co smoke. Seminars in Oncology (1976). 29. Wynder, E. L.. Mabuchi. K., and B.~attie, E. J. The epidernioiogy of lung cancer: Recent trends. 1. Amer. Med. Assn. 213, 2221-222g 41970), 30. Wynder, E. L.. and SteIlman. S. D. hnl~Ct of long-term tilter cigarette u~age on lung and laryn~ r, an~er risk: ^ case-control study. J. Natl. Cancer lest. 62, 471-477 (1979). 31. Yuspa, $. H.. and Harris, C. G. Molecular ~-nd cellular basis of chemical caxcinogenesi~, in "'Cancer Epidemiol?gy and Prevention" (D. Szhottenfeld and J. F. Fraumeni, Eds.), pp. 23-43. W. B. Saundets, Philadelphia, 1982. 32. Ziegler, R.. Mason, T. L, Stemhagen, A., Hoover. R., 5¢hoenberg, J., Gridl¢~, G., Virgo. P., and Fraumerd, J. F, Carotenoid intake, vegetables, and the risk of lung cancer among white men in N~w Jersey. Amer. J. Epldemioi. 123, 1080-1093 (1985).