Tobacco Documents Online

WORLD HEALTH ORGANIZATION INTERNATIONAL AGENCY FOR RESEARCH ON CANCER A TOBACCO: MAJOR INTERNATIONAL HEALTH HAZARD Proceedings of an International Meeting organized by the IARC and co-sponsored by the All-Union Cancer Research Centre of the Academy of Medical Sciences of the USSR, Moscow, USSR held in Moscow, 4-6 June 1985 EDITORS D. G. ZARIDZE R. PETO IARC Scientific Publications No. 74 INTERNATIONAL AGENCY FOR RESEARCH ON CANCER LYON 1986

CIGARETTE YIELD AND CANCER RISK: EVIDENCE FROM CASE-CONTROL AND PROSPECTIVE STUDIES S.D. STELLMAN American Cancer Society, Inc. 4 West 35th Street, New York, NY 10001, USA INTRODUCTION The belief that cancer risk can be reduced by lowering the tar yield of cigarettes has been developed from three basic observations: (1) many cancers exhibit a dose-response with respect to the number of cigarettes smoked per day, as shown in Figure 1 (Wynder & Stellman, 1977); (2) cancer risk decreases with number of years of smoking cessation (Fig. 2); (3) tumours can be produced quantitatively in animals using cigarette combustion products (Wynder & Hoffmann, 1967). Although quantitative relationships between cigarette smoking and cancer risk had been developed in both case-control and prospective studies in the 1950s and even earlier, epidemiological confirmation of a specific relationship with cigarette tar yield was not achieved consistently until the late 1960s. Since that time, differences in relative risk have been observed for at least four cancer sites: lung, larynx, oral cavity, and bladder. In this paper we review the data which have led to these conclusions, and discuss some of the similarities and differences in the studies. LUNG CANCER Case-control studies Three series of case-control studies have estimated the relative risk for developing lung cancer in relation to cigarette yield: Bross and Gibson (1968), the series begun by Wynder in the 1960s and continuing into the present (Wynder et al., 1970; Wynder et al., 1976; Wynder & Goldsmith, 1977; Wynder & Stetlman, 1977; Mushinski & Stellman, 1978; Wynder & Stellman, 1979; Wynder et al., 1984), and a cooperative European study begun in 1976 under the auspices of the US National Cancer Institute, covering five countries: the results have been presented as a whole (Lubin et al., 1984a,b) and the Austrian - 197- 03

198 STELLMAN Fig. 1. Relative risk for cancers of the lung (Kreyberg types ~ and II), oral cavity, larynx, oesophagus, and bladder for male current smokers, according to number of cigarettes smoked per day. N, number of cases in case-control • study (from Wynder & Stellman, 1977) lOOI LUNG CANCER (Z) 251 LUNG CANCER N : 486 CASES 1 20 ~ 604 ~ 20 5 ~ O' 0 ~ ORAL CAWTY LARYNX o N: 388 N~27C ~ 20 SMOKER NON- ~-~3 II-ZO 2~-30 3~-~,O 4~* SMOKER OESOR~GUS N:I2~ BLADDER N= 384 NON- I-I0 ll-20 21-30 31-40 41+ NON- I-IO II-20 21-30 31-40 41+ SMOKERSMOKER NO OF CIGARETTES SMOKED PER DAY component has also been published separately (Kurtze & Vutuc, 1980; Vutuc & Kunze, 1982a,b, 1983). Results of these case-control studies are summarized in Table 1, in which comparisons are made between smokers of filter versus nonfilter cigarettes. The relative risk of lung cancer in nonfilter as comoared to filter cigarette smokers as a referent rnn~e~ from 1 3 to 2.3. This must be understood in the context of an individual's lifetime exposure to cigarette tar. The average age of lung cancer diagnosis in the USA is now about 58 years. Most

ClGARElq-E YIELD AND CANCER RISK 199 Fig. 2. Relative risk for cancers of the lung (Kreyberg types I and ll), oral cavity, larynx, oesophagus, and bladder for male former cigarette smokers, according to number of years since cessation of smoking. N, number of cases in case-control study (from Wynder & Stellman, 1977) 16! LUNG CANCER LUNG CANCER(I) CASES=687 CONTROLS=6534 YEARS OF SMOKING CESSATION smokers in this cohort began smoking at a time when there were very few filter cigarettes on the market, and the tar yield of nonfilter cigarettes was over 30 rag. Data from the new Ameriean Cancer Society study (Stellman & Garfinkel, 1986) suggest that a wave of switching from nonfilter to filter cigarettes occurred in the mid-1960s immediately after the appearance of the Surgeon General's report in 1964, which received widespread publicity. Figure 3 shows the proportion of a smoker's lifetime which would have been spent with filter cigarettes, assuming smokers switched from nonfilter cigarettes at about that time, and assuming average ages of beginning to smoke characteristic of this population. It is obvious that recent lung cancer cases received a great deal of their tar exposure in their early smoking years from nonfilter, or from the early high-tar filter cigarettes, irrespective of the types of cigarette they smoke today.

200 STELLMAN Table 1. Relative risks for lung cancer reported from case-control studies, in relation to filter usage' Study Sex Comparison Relative risk Bross & Gibson (1968) Males Fto NSR 3.8 NFto NSR 6.5 NF to F 1.7 Wynder etal.(1970) ~ Males F to NSR 23.6 NFto NSR 38.3 NF to F 1.6 Wynder & Stetlman (1979) Males NF to LTF 1.3 Females NF to LTF 1.4 Lubin et al. (1984 a, b) Males Mixed F and NF to F 2.1 NFto F 2.I Females Mixed Fand NFto F 2.3 NF to F 2.3 • Abbreviations: F, filter cigarette smokers; NSR, nonsmokers; NF, nonfilter cigarette smokers; LTF, long-term fiiter cigarette smokers ~Cases were Kreyberg type I only In three of these case-control series, results have also been presented in terms of specific tar yields. These findings, shown in Table 2, demonstrate that, even allowing for substan- tial differences in schemes for estimating smokers' tar dosage, dose-response relationships are easily discernible. Follow-up studies There have been three important follow-up studies of lung cancer in relation to cigarette smoking in which cigarette yield has been studied in detail The American Cancer Society enrolled over one million men and women aged 40 years and over, in 25 states, in a prospective study in 1959. Follow-ups were conducted annually through 1966, and again in t971 and 1972. Analyses of lung cancer death rates in relation to smoking habits were originally published by Hammond (1966). Hammond et al. (1976, 1977) later presented evidence from this study showing that the lung cancer mortality rates for smokers of 'low tar-nicotine' cigarettes, compared to rates in smokers of 'high tar-nicotine' cigarettes, were reduced by about 20% in men and by about 40% in women. These estimates were made using a matched group analysis which permitted adjustment for many variables at once, including age, race, number of cigarettes smoked per dav. a~e smokin~ began, urban/rural residence, education, iob exoosure to chemicals, X-rays, or other toxicants, history of prior illness, and calendar period (Ham- mond, 1985). Hammond's results are shown in Table 3. For the present review we have re-calculated the standard mortality ratios (SMR) according to quantity smoked daily by current smokers, and by tar yield of cigarette at baseline, for lung cancer in men during 1960-1966, the six years when annual follow-up was done. Calculations were also restricted to this period to minimize effects of changes in

CIGARE'TqE YIELD AND CANCER RISK 201 Fig. 3. Filter cigarette usage as a percentage of total smoking experience, by birth cohort (from Wynder & Stellman, 1979) 70, 60. 50- 20- 10- ~FI LTER i INON-FILTER 25% 5o% 58% 5O% 75% YEAR OF BIRTH smoking habits. In addition, during the first six years of the study, additional confirmation was sought whenever cancer was mentioned on the death certificate, so that the cause of death was based upon 'best evidence'. "Results of this new calculation are shown in Figure 4. There were 967 deaths from lung cancer during this period. For statistical convenience, the reference population is the ,4,° ......e,~,~'v, ~-,~-'~, ~L~CL~ U~ m=diura ~ar-nicoane ctgare~tes, wt~o smoked 20 cigarettes per day. For all other tar-nicotine and quantity categories of smokers, as well as for exsmokers and nonsmokers, expected numbers of deaths were computed by multiply- ing age-calendar-year-specific lung cancer death rates in the reference population by the person-years of exposure to risk of dying in the target group, and summing over age- calendar-year strata. The SMR is the number of observed divided by e:epected deaths. Data were renormalized to give lifetime nonsmokers an SMR of 1.0. 0 O~ O~ 0

202 STELLMAN Table 2. Relative risk for lung cancer according to tar exposure indices proposed by various authors" Reference Sex Relative risk Mushinski & Stellman (1978) Current tar level (rag/day) 0 1- 200- 400- 600- 800- 199 399 599 799 999 1000-1200-1400+ 1199 1399 Kreyberg I Males1.0 5.1 7.4 12.2 20.1 24.8 34.2 30.6 29.9 Females 1.0 7.9 9.6 18.9 28.5 14.8 Kunze & Vutuc (1980); Lifetime tar score Vutuc&Kunze(1982b) Below 501- 1001- 2001- 3001+ 500 1000 2000 3000 Kreyberg t Males2.0 2.6 5.3 7.2 8.3 Females 1.5 4.2 4.8 4.9 6.8 Kreyberg II Males - 1.8 1.8 3.5 3.9 Females - 1.1 3.1 - 2.3 Lubin etal, (1984a) Mean cigarette tar content (mg) ~ (15.6) (18.5) (20.6) (23.6) (25.2) (28.8) Lung cancer Males 1.0 1.2 1.7 1.3 1.3 1.4 Females 1.0 1,9 1.3 1.1 1.5 - Nonsmokers and referent; see Table 5 for definitions of tar exposure indices Categories were combined from wRhin-country 10, 25, 50, 75, and 90th percentiles. Mean tar values (given in brackets) are within each such category Table 3. Standardized mortality ratio for lung cancer among one million men and women followed up for twelve years, relative to lifetime nonsmokers, according to tar-nicotine yield of usual cigarettes, adjusted for age, calendar year, and many other variables (see text) = Standardized mortality ratio 'Low T/N' 'Medium TIN' Males 0.81 0.95 1.00 Females 0.60 0.79 1.00 'Adjusted' deaths: 235.2 285.5 318.4 mFrom Hammond et aL (1976}

CIGARE-i-rE YIELD AND CANCER RISK 203 Fig. 4. Standardized mortality ratios for lung cancer in males, among nonsmokers, exsmokers, and current smokers of low-, medium-, and high-tar/nicotine (T/N) cigarettes (defined by Hammond etaL, 1976). The group was enrolled in 1959, and followed up through 1966. 2O ~X-~,~ i....,.- e . ..... MEDIUM SMOKER S~,......'"" • ........... LOW ~ ./NON-SMOKERS l I L I t t 0 I0 ZO ~0 4.0 4.5 CIGARETTES SMOKED PER DAY At each tar-nicotine level, the SMR increased with quantity smoked, in an approxi- mately linear dose-response relationship. For current smokers of at least 20 cigarettes per day, at each value of daily quantity smoked, the SMR for the 'high tar-nicotine' cigarette smokers exceeded that for the 'medium' group, which in turn exceeded that for the 'low' group. Lifetime non-smokers had lung cancer death rates well below any of the current smokers, irrespective of cigarette yield for the latter. Two other studies are worthy of mention. Rimington (1981) observed 104 lung cancer cases in a follow-up study of 10 414 male volunteers for a mass radiography screening in England. Subjects were enrolled in 1970--1971, and followed for 69 to 81 months. The relative risk for nonfilter versus filter cigarette smokers was reported as 1.54. The inci- dence was computed by dividing the numbers of cases by numbers enrolled, without considering person-years at risk. Adjustment was made for age and for quantity smoked. .In the Whitehall study (Higenbottam et al., 1982), smoking data were available for 17 475 of 18 403 male civil servants aged 40-64 years who were enrolled during 196%1969 and followed for at least ten vears. Ten-~'o~ ~,-~t~ ~-~*~,~ ~.4~,_,_~ted f~r age a=~ cm~Ic3-r-cnt grade, were computed for current smo'kers within categories of inhalation, quantity and tar-yield. There were 108 deaths due to lung cancer among inhalers, and 35 among noninhalers, with tar- and quantity-specific rates for both groups shown in Table 4. Among inhalers, the data show a distinct dose-response at the two lowest consumption levels (1-9 and 10-19 cigarettes per day), although not at the highest, and among noninhalers there is a possible dose-response at the two highest levels (10-19 and 20 or more cigarettes per day). t li;. II tiii!! OTM O~ O~ O

2O4 STELLMAN Table 4. Ten-year lung cancermortality rates (and numberofdeaths) among 17 475 male British civil servants in the Whitehall study, according to quantity smoked, tar yield, and inhalationm No. cigarettes smoked per day Tar yield (rag) 1~3 2~2 ~+ Inhalers 1-9 0.39 (2) 0.53 (1) 1.62 (7) 10-19 1.46(19) 1.55 (8) 2.61 (20) 20+ 2.23 (35) 2.00 (13) 1.79 (3) Noninhalers 1---9 1.08 (4) 0.00 (0) 0.93 (1) 1 0-19 1.25 (5) 1.28 (2) 4.18 (5) 20+ 1.71 (7) 5.81 (9) 5.85 (2) • From Higenbottam et aL (1982) ..2_ co ci: ca 1o ca re ca st: at CANCERS OTHER THA~N LUNG Studies of cigarette yield and cancer have focused mainly on lung cancer, for the obvious reason that, having the greatest incidence and mortality rate of tobacco-related cancers, the numbers of cases available for study are greater than for other sites. Several studies, however, have examined the possible influence of cigarette yield on other cancers. In the American Health Foundation case-control studies, interviewers were instructed to see patients with cancers of the tung, mouth, oesophagus, larynx and bladder. Wynder and Stellman (1979) published relative risks for cancer of the larynx based on 286 male and 64 female cases. After adjusting for age, duration of smoking, number of cigarettes per day, and alcohol consumption, the risk of larynx cancer in nonfilter versus long-term filter cigarette smokers (at least ten years on filters) was 1.49 for men and 3.97 for women (both significant). The relative risk was greater for nonfilter than for filter cigarette smokers at every quantity level. Lee and Garfinkel (1981) reported new analyses of data from the American Cancer Society follow-up study of 1959-1972, in which the relative mortality for smokers of low tar/nicotine cigarettes (as defined by Hammond et aL, 1976) was consistently lower in both men and women than for high tar/nicotine cigarettes for cancer of the buccal cavity and pharynx, oesophagus, larynx, bladder and pancreas. The adjustment procedure, based upon simultaneous matching for nine separate variables, rendered the numbers of effec- tive ('adjusted') cases very small. The mortality rati0g were statistically significant only for cancers of the oesophagus and bladder in women, and for none of the sites in men. Wynder et al. (1976) gave relative risks for cancer of the oral cavity in a case-control study of 593 men and 280 women and matched controls: for nonfilter cigarette smokers versus nonsmokers, 7.8; for long-term filter cigarette smokers versus nonsmokers, 5.7; and for nonfilter versus long-term filter cigarette smokers, 1.4. Adjustment was made for age, but not for alcohol consumption. Significance levels were not given.

CIGARER-E YIELD AND CANCER RISK 205 In a Canadian, population-based, case-control study of 480 male and 152 female case- control pairs, Howe et aI. (1980) reported a reduced risk associated with the use of filter cigarettes compared to nonfilter cigarettes. A recent Italian study of 512 male bladder cancer cases and 596 controls gave a relative risk of 3.0 for nonfilter versus filter cigarette smokers (Vineis e~ al., 1984). On the other hand, there was no difference for men between long-term filter and nonfilter cigarette smokers in the relative risk for bladder cancer in a case-control study by Wynder and Goldsmith (1977), which involved 574 cases and an equal number of matched controls. DISCUSSION There are many methodological issues which must be dealt with in the assessment of the relationship between cigarette yield and cancer outcomes. These fall roughly into four categories: questions of dosage, outcome, other etiological factors and confounding. The strengths and weaknesses of the studies described may be examined largely through attention to these four items. Dosage In any study of cigarette type and disease, dosage is the most important - and in some ways the most difficult - variable to estimate. There are many reasons for this. In the first place, the average tar content of cigarettes has fallen considerably during the past 30 years, even within the same brand. Secondly, some smokers switch brands fre- quently, particularly in response to promotion of the new brands or in response to 'health' publicity. Thirdly, most smokers try to quit at some time in their lives; some are successful, others quit and begin again repeatedly. The actual lifetime dosage of persons in the latter category is quite difficult to determine. Finally, even in well-conducted interviews, sub- jects sometimes recall their smoking history imperfectly, especially regarding duration of smoking specific brands. Many different ways of expressing cigarette dosage have been used, ranging from simple classification as filter versus nonfilter, to elaborate algorithms designed to account for 'complete' year-by-year smoking histories. Cumulative dosage measures have the advan- tage of taking into account the subject's entire history, including early smoking, which may have contributed disproportionately to lifetime tar exposure, since the cigarettes first smoked by persons now in the cancer age group had tar contents two to three times those of current cigarettes. It has the disadvantage of making cumulative scores 'pile up' at the be~innin~ of a smoker's life. during the ve~r~ scores may be insensitive to differences in tar levels between recent brands. Furthermore, cumulative dosage scores, particularly when expressed as 'pack-years', have the disadvan- tage of making two packs per day for 10 years equivalent to one pack per day for20 years, necessitating further adjustment for duration or other parameters. The wide range of tar exposure indices which have been used by various authors is shown in Table 5. These range from categorization of smokers as either filter or nonfitter cigarette smokers (Bross & Gibson, 1968; Wynder & Stellman, 1979), use of the tar rating of the