0 P z : " 0 N N RRTES/lOO.OOO FIGURE 4.-Age-specific mortality rates for nonwhites in the United States for cancer of the bronchus, trachea, and lung SOURCE: Nst~onal Cancer Institute (1981. 26 I I I I 1 IIIIIIIIIII llll~llllllllllll~ ~llIIIIlIIIIIIlI Id IGO ?IGURE B.-Age-specific mortality rates by &year age groups for cancer of the bronchus, trachea, and lung for white males, United States, 1950- 1977 SOURCE: Natronal Cancer Institute ,198) The term "lung cancer" refers to a number of specific malignant iseases involving the lungs. Several systems of classifying lung ancer have been proposed (Table 1). Four cell types constitute the majority of lung cancers: epidermoid r squamous, adenocarcinoma, small cell (oat cell), and large cell. `here are differences in the frequency distribution of the different 27 377-310 0 - 82 - 4 FIGURE 6.-Age-specific mortality rates by B-year age groups for cancer of the bronchus, trachea, and lung for white females, United States, 1950-1977 SOURCE Natmnal (`ancer Instltute~ 198IHI types of lung cancer in males and females and in smokers and nonsmokers. Epidermoid carcinoma was the most common histologi- cal type of lung cancer in the male smoker, while adenocarcinoma was most common in the female smoker and in nonsmokers of both sexes in a series recently published from the Mayo Clinic (Table 2) (225). Other centers have reported similar data, although the 28 FABLE l.-Comparison of the World Health Organization (WHO), Veterans Administration Lung Cancer Chemotherapy Study Group (VALG), and Working Party for Therapy of Lung Cancer (WP-L) Lung Cancer Classifications WHO -_ I. Epidermoid carcinoma VALG WP-L 11' Small cell carcincma 1. Fusiform 2. Polygonal 1. Squamous cell carcinoma 10. Epidermoid carcinoma a. With abundant keratin 11. Well differentiated b. With intercellular bridges 12. Moderately differentiated c Without keratin or 13. Poorly differentiated bridges 2. Sma:! cell carcinoma 20. Small cell carcinoma a With oatcell structure 21. Lymphocytelike h. With polygonal cell 22. Intermediate cell structure 3. Lymphocytelike 4. Others III. Adenocarcinoma 1. Bronchogenic a. Acinar h. Papillary 2. Bronchoalveolar IV. Large cell carcinoma 1. Solid tuaor with mucin 2. Solid tumor without mucin 3. Giant cell 4. Clear cell 3. Adenwarcinoma 30. Adenocarcinoma a. Acinar 31 Well differentiated b. Papillary 32. hloderately differentiated c. Poorly differentiated 33. Poorly differentiated 34. Bronchiolopapillary 4. Iarge cell undifferentiated 40. Large cell carcinoma 41. With stratification 42. Giant cell 43. With mu& formation 44. Clear cell SOURCE: Matthews and Gordon (176). proportions by histological type vary with the pathological criteria used, the patient population, the geographic location, and other factors. Earlier epidemiologic studies suggested that cigarette smok- ers were more likely to develop squamous cell, large cell, and small cell lung carcinoma than other types (67, 148). This view has been supported by some investigators (54, 284) and disputed by others (6, 18, 19, 137, 293. 329). More recent investigations indicate that all four major histological types of lung cancer-including adenocarci- noma, which appears to be increasing in recent years-are related to cigarette smoking in both males and females (8, 284, 293). Establishment of the Association Between Smoking and Lung Cancer It is not ethical or feasible to perform a controlled experiment in humans to establish a causal relationship between tobacco smoking and lung cancer. Practically, epidemiological methods are employed to test a causal hypothesis. These methods, as discussed previously, when coupled with pathological and experimental data, provide the framework for a judgment of causality. 29 TABLE Z.-Histologic types of pulmonary cancers in smokers and nonsmokers Type TOtal Smokers Smokers Epidermoid 992 892 7 80 13 Small cell 640 533 4 100 3 Adenocarcinoma 760 492 39 128 101 Large cell 466 389 16 46 15 Bronchi&alveolar 68 35 4 13 16 TOtal 2,926 2,341 70 367 148 SOURCE: Rosenow 1225) Numerous retrospective studies have examined smoking patterns among established cases of lung cancer and a variety of matched controls. These studies have been summarized and reviewed in previous reports from the Department of Health and Human Services (270,272-281). Eight prospective studies have measured lung cancer mortality rates among smokers and nonsmokers followed over various time intervals. In October 1951, Doll and Hill (62, 63) initiated the first major prospective study of the relationship between smoking habits and mortality in a cohort of more than 40,000 male and female physicians. By 1965, seven other major prospective studies in four countries had been initiated. These studies cumulatively represent more than 17 million person-years of observation and over 330,000 deaths. The study designs are summarized below and in Table 3. The number of years of followup reported for the various major prospective studies ranges from a low of 4 years in the American Cancer Society Nine-State Study to 22 years for females in the British Physicians Study. Published reports for the varying followup periods differ substantially for each study with respect to the amount of information provided. Data from the Japanese study have been published presenting 5, 8, 10, and 13 years' results. For each followup period, site-specific cancer mortality is fragmented. Data for specific cancer sites are available only for males from the 13-year followup study; dosage analyses for other cancer sites for either males or females are intermittent among the many published reports cited. In all cases, the most current data from each of the prospective investigations are cited. In some instances, mortality rates (or ratios) for all smokers for a specific site may be from one study period while dosage information (usually expressed as the number of cigarettes smoked per day) may be from another (followup) period. The reader is referred to the references cited at the end of each study description for a complete bibliography. 30 The British Physicians Study In.1951, the British Medical Association forwarded to all British doctor% a questionnaire about their smoking habits. A total of 34,400 men and 6,207 women responded. With few exceptions, all physi- cians who replied in 1951 were followed to their deaths or for a minimum of 20 years (males) or 22 years (females). Further inquiries about changes in tobacco use and some additional demographic characteristics of the men were made in 1957,1966, and 1972 and of the women in 1961 and 1973. By 1973 more than 11,000 deaths from all causes had occurred in this population (62-66, 68, 69, 71). The American Cancer Society 25State Study In late 1959 and early 1960, the American Cancer Society enrolled 1,078,894 men and women in a prospective st.udy (97-102, 155). Although this was not a representative sample of the United States population, all segments of the population were included except groups that the planners believed could not be traced easily. An initial questionnaire was administered that contained information on age, sex, race, education, place of residence, family history, past diseases, present physical complaints, occupational exposures, and various habits. Information on smoking included type of tobacco used, number of cigarettes smoked per day, inhalation, age started smoking, and the brand of cigarettes used. Nearly 93 percent of the survivors were successfully followed for a la-year period. Early reports of this study examined lung cancer mortality in relationship to several parameters of smoke exposure, including duration of habit and age at onset, among others. Two recent reports have examined the effects of general air pollution (101), the type of cigarette smoked (155), and lung cancer mortality. Cancer mortality data for 483,000 white females and 358,006 white males for the period 1967 to 1971 were also recently reported (106). The U.S. Veterans Study The U.S. Veterans study (74, 131, 222-224) followed the mortality experience of 290,000 U.S. veterans who held government, life insurance policies in December 1953. Almost all policyholders were white males. The data for specific causes of death during a 16year period were recently reported by Rogot (224) and are similar to earlier data published after only S'/, years of observation of this population (131). Over 107,000 deaths have occurred in this popula- tion. The Japanese Study of 29 Health Districts In late 1965, a total of 265,118 men and women in 29 districts in Japan were enrolled in a prospective study (115-120). This represent- 31 ed from 91 to 99 percent of the population aged 40 and older in these districts. This study provided the unique opportunity to examine the relationship of cigarette smoking to death rates in a population with genetic, dietary, and cultural differences from previously examined Western populations. By the end of the 13th year of followup, almost 40,000 deaths had occurred, including 10,300 cancer deaths, and there were over 3,000,OOO person-years of observation. For females, the main body of published data is based on 5 to 8 years of followup. The Canadian Veterans Study Beginning in 1955, the Canadian Department of National Health and Welfare enrolled 78,000 men and 14,000 women in a study of smoking-related mortality (26, 27). Information was obtained on age, detailed smoking history, residence, and occupation. During the first 6 years of followup, 9,491 males and 1,794 females died. No more recent followup has been reported. The American Cancer Society Nine-State Study In the American Cancer Society Nine-State Study (104, 105), 187,783 white males were followed for an average of 44 months. This study began in early 1952. There were 11,870 deaths in the age 50 to 70 population. The last major report of this study was published in 1958. The California Men in Various Occupations Study This study (76, 290) examined the mortality experience of 68,153 men, 35 to 64 years of age, over a period of 482,650 person-years of observation. A total of 4,706 deaths occurred. These men were in nine occupational groups. The last published report from this study was in 1970. The Swedish Study A national probability sample (42) of 55,000 Swedish men and women was surveyed in 1963 by mailed questionnaires, to which 89 percent of the sample responded. Information was collected on smoking status at the time of the initial query and for specific intervals during the previous 9 years according to type and amount of smoking and degree of inhalation. The questionnaire identified age, sex, location (urban, nonurban), income, and occupation of subjects. A lo-year followup on smoking-related mortality was published in 1975. 32 TABLE 3.-Outlin9 of eight major prospective studies Doll Dorn Best Weir cederlof Authors Hill Hammond Kahn Himyama Joeie Hammond Dunn fiberg Pet0 Ronot Walker Horn Linden Hrubec Pike Breelow lmich Males end Total population California Probability British fern&e U.S. of Canadian White malee sample of Subjects in 29 health males in doctors in the 25 veterans districts in various pensioners nine states Swedish States Japan urupetiona population Population size w@o WW@J 2%m -2%~ 92.~ 187.ou) woo %ooo Females 6,~ 562,671 1.78 13.06 17.00 > 3.70 Comments Swedish Study Nonsmoker None Light DeeP 1.00 1.00 Female data 3.70 - based on only 7.80 7.20 9 total lung 9.20 .l.SO cancer deaths Temporal Relationship of the Association The criterion of temporality requires that cigarette smoking antedate the onset of cancer. Suppdrt for this criterion is provided by all the major prospective studies in which an enormous number of initially disease-free subjects were followed over varying time intervals. 39 LUNG CANCER I. MALES 80 70 60 xl 40 30 20 10 N: CASES CONTROLS NON F NF F NF F NF F NF F NF SMOKER l-10 11-20 21-30 3140 41+ NO. OF CtGARETTES SMOKED PER DAY xl 25 126 iz FIGURE 7.-Relative risk of lung cancer for males, by number of cigarettes smoked per day and long-term use of filter (F) or nonfilter (NF) cigarettes SOURCE: W'ynder (3271 Indirect support for the temporality of the association is provided by other studies (57, 70). One study (57) examined the relationship between per capita tobacco consumption in 1930 and male lung cancer death rates in 1950 in I.1 different countries (Figure 9). This study encompassed the era prior to the advent of filter cigarettes. Assuming that the majority of tobacco consumption in 1930 occurred among males and that there was a 20-year latency period for the development of lung cancer, there was a strong positive correlation between tobacco consumption in 1930 and lung cancer death rates in 1950. 40 LUNG CANCER I. FEMALES CASES N: CONTROLS NON F NF SMOKER l-10 F NF F NF F NF 11-20 21-30 31+ NO. OF CIGARETTES SMOKED PER DAY FIGURE 8.-Relative risk of lung cancer for females, by number of cigarettes smoked per day and long-term use of filter (F) and nonfilter (NF) cigarettes SOURCE: Wynder (327). A later study (70) examined the relationship between manufac- tured cigarette consumption per adult in 1950 and lung cancer death rates in males and females who were in the 35- to 44-year-old age group in the mid-1970s (who had entered adult life in 1950). There Was a consistent correlation between cigarette consumption and lung cancer death rates in different countries (Figure IO), a finding which WaS "better than...expected in view of the possible international differences in cigarette composition, puff frequency, style of inhala- tion, butt length, additional use of nonmanufactured cigarettes (and other forms of tobacco), and national consumption of cigarettes in intervening years between 1950 and 1975." 41 TABLE 9.-Age-adjusted lung cancer mortality ratios for males and females, by tar and nicotine in cigarettes smoked M&S Females - -. High T/N 1.00 1.00 \ Medium T/S 0.95 0.79 Low T/N 0.81 0.60 the *The mortality rstm for the category with highest risk was made 1.00 90 that the relative reductions in fia nJ use of lower T/N cigarettes could be visualiud. SOURCE: Hammond et al. c 1031 Additional evidence for the temporality of this association b advanced by a number of histological studies showing that smoke& develop histologic changes interpreted by most pathologists a premalignant lesions in bronchial epithelium in much greater proportions than nonsmokers, and that these changes progra toward cancer in continuing smokers but reverse in ex-smokers (9 14, 15) (Table 14). Coherence of the Association The final criterion is the coherence of the association betwwc smoking and lung cancer with known facts in the biology and natural history of lung cancer. Coherence of the association has been noted with the following facts: Dose-Response Relationship Between Smoking and Lung Cancer Mortality The finding of a dose-response relationship between cigarette smoking and lung cancer provides great coherence with the known facts of the disease. Regardless of the measure of tobacco consump tion employed (i.e., number of cigarettes smoked, inhalation practice, duration of smoking, age when smoking began, or type of cigarettes smoked), there was a gradierut of disease consistent with a true dose response relationship in ever:y study. Sex Differences in Lung Cancer Mortality Correlating With Corresponding Differences in Smoking Habits Males have had higher lung cancer death rates than females. This observation has been interpreted by some as contradictory to the causal role of smoking in lung cancer (8.2, 167). However, a careful examination of smoking patterns and age-specific mortality data ha 42 GREAT BkTAIN # I I I I I CIGARETTE CONSUMPTION FIGURE 9.-Crude male death rate for lung cancer in 1950 and per capita consumption of cigarettes in 1930 in various countries SOURCE. DOII 1.57, been interpreted by most observers as support for the causality of smoking in lung cancer. Historically, males began to smoke in large numbers in the World War I period, and much of the increased cigarette use noted during this period reflected switching from other forms of tobacco (e.g., smokeless tobaccos, pipes, and cigars) to cigarettes. Females began to smoke in larger numbers about 20 to 25 Years later, in the World War II era (270); at that time, a smaller Proportion of females smoked compared to males, and those who did, generally smoked fewer cigarettes per day, inhaled less, started later in life, and were more likely to smoke lower tar and nicotine and filtered cigarettes. These differences in smoking habits of males and 43 377-310 0 - 82 - 5 a Ratesbasedcmow100~~ 0 Rateabased0n25-1OOdeah 0 U.S. non-smokers 19%197p SMOKED 0` 1 t 1 1 500 1000 lWC! 2ooo 2500 3aoo MANUFACTUREIY ClGARElTES PER AWLT IN 1950 FIGURE lo.-International correlation between manufactured cigarette consumption per adult in 1959 while one particular generation was entering adult life (in 19501, and lung cancer rates in that generation as it enters middle age (in the mid-i97Os) NOTE Comparison has been restricted to developed countries (i.e.. excluding Africa, all of Asia except Japan. and all except North Amencsl. with populations > 1 million, Lo impmve the accuracy of the &sewed death certification rates aa indicators of the underlying riskn of lung cancer among people aged 3&44. `Lung cancer death certification rates per million. adults aged 3544 are from WHO 003,304l These rates are the means of the male and female rates for all yews (1973.1974. or 1975) reported in WHO (303). except for Greece (which was not reported in WHO (3031 and thus was taken from WHO (Xl0 and Norway for which the rates in WHO 130.X and WHO 1304I were based on only 11 and 14 cases, respectively; for statistical stability, these welp averaged. "Manufactured cigarettes per adult are from Lee (Jw for the year 1950 Iexcept for Italy. where consumption data are available in 5-year groups onlyl; to avoid tl'le temporary postwar shortages. data for 1951-55 have been used. This excludes handrolled cigarettes, which in most countries accounted for only a small fraction of all cigarette tobacco in 1950. `U.S. nonsmoker rates were estimated by fitting straight lines (on a double logarithmic eeale) to the relationship between lung cancer mortality and age reported for male and for female lifelong nonsmokers by Gartinkel(86) and averaging the plpdicted values at age 40. (Although the average of the male and female rati actually observed at these ages is sinular to this estimated value. thew observed rates are each based on fewer than five c~dc8 tGarfinkell(86l and so might have been inaccurate.l SOURCE: Doll and Peto( 70). females correlate well with the observed sex differences in lung cancer mortality rates. In fact, the rise in female lung cancer mortality rates observed in the late 1950s and early 1960s appears to be reproducing the phenomena noted among males 20 to 30 years earlier. If one subtracts 25 years from the female cancer death rate, as noted previously in Figure 1, the rates for women are only slightly below the rates for men. Thus, close scrutiny of these trends reveals 44 no substantial difference in the risk of developing lung cancer between men and women. Lung Cancer Mortality and Cessation of Smoking Since cigarette smoking is significantly associated with lung cancer, it is logical to expect that cessation of smoking would lead to a decrease in mortality rates from lung cancer among quitters compared to persons who continue to smoke cigarettes. In fact, all of the major studies which examined cessation showed this decrease in lung cancer risk. Data from four of the major prospective studies are presented in Table 10 for illustration. After 15 to 20 years, the ex- smoker's risk of dying from lung cancer gradually decreases to a point where it more closely approximates the risk of the nonsmoker (68, 224), whereas for the continuing cigarette smoker, the lung cancer risk is more than 10 times that of the nonsmoker. The magnitude of the residual risk that ex-smokers experience is largely determined by the cumulative exposure to tobacco prior to smoking cessation (i.e., total amount the individual smoked, age when smoking began, and degree of inhalation), and varies with number of years since quitting smoking, as well as with the reasons for quitting smoking (e.g., quitting due to symptoms of disease). Differences in Lung Cancer Mortality by Site of Residence (Urban Versus Rural) A number of studies have examined the relationship of smoking to lung cancer mortality by site of residence (urban or rural) and air quality of a community. Eight of the earlier studies were reviewed in the 1971 Report of the Surgeon General (276). More recent publica- tions include "Epidemiological Review of Lung Cancer in Man" (111) and the report of a task group, "Air Pollution and Cancer" (41). There have been studies in England and Wales (59), in 20 countries combined (40, 291), as well as in the United States (101, 146, 164, 258). The majority of these studies has found that lung cancer mortality is more common in urban than rural areas. This urban to rural gradient is primarily, but not exclusively, found among smokers. Since cigarette consumption is generally greater in urban areas than in rural areas, it is difficult to define conclusively what proportion, if any, of the excess lung cancer mortality in city dwellers can be accounted for by urban living independent of smoking. One study (164) examined the risk of several cancers by religion and place of residence in 20,379 cases in the State of Utah. Members of the Church of Jesus Christ of Latter-Day Saints (Mormons) composed approximately 70 percent of the state's population in 1970. The use of tobacco and alcohol is prohibited by religious tenets, and it is documented that Mormons have a very low proportion of 45 TABLE IO.-Lung cancer mortality ratios in ex-cigarette smokers, by number of years stopped smoking U.S. Veterans ' `Years stopped Study smoking Mortality ratio British Physicians 14 16.0 5-9 5.9 10-14 5.3 15 + 2.0 Current smokers 14.0 l-4 18.83 5-9 7.73 lo-14 4.71 15-1s 4.81 20+ 2.10 Current smokers 11.28 l-4 4.65 5-9 2.56 10 + 1.35 Current smokers 3.76 ACS 2.5-Stat.e Study (males 50-69)