TABLE l.-Detailed computation of smoking-attributable lung cancer deaths among females, United States, 1985 Exposure category - Prevalence Relative riska Assigned share Case fraction p ml r s (%`o) f(S) Current smokers I-10perdayh II-19 per day 20 per day 2 I-30 per day 23 1 per day Former smokers C&2 yearsC 3-5 years 6 10 years I I-15 years 2 I6 years 9.3 5.5 81.9 9.4 3.3 11.3 91.1 6.7 9.3 14.2 93.0 24.0 3.2 20.4 95.1 11.8 2.7 22.3 95.5 10.8 5.0 18.2 94.1 16.7 2.5 I I.2 91.1 5.0 3.4 4.9 19.5 3.0 2.0 3.2 68.5 1.2 4.0 1.8 43.4 1.3 Exposure Prevalence category p (%J) Case fractions Attributable Attributagle A%) risk deaths a (%) A Current smokers 27.8 62.7 57.7 22,300 Former smokers 16.9 21.2 24. I 9,300 Current and former smokers 44.7 89.9 81.8 31,600 "Ratlo of age-adJusted death rates. where age adJustment was performed by d!rect amdardzarion to the age di$tnbution of woman-yean of exposure among nonsmokers. hNumtxr of cigarettes smoked per day. as of the date of enrollment (September 1982). `Number of years elapsed since last smoked regularly. as of the date of enrollment (September 1982). `Attnbutable deaths A equal oD. where u I\ attribuuble risk and D equal\ 38,687 lung cancer deaths among adult females in 1985. SOURCE: Gafinkel and Stellman (1988): THIS 1985. unpublished tabulatmns: NCHS, Dtvtsion of Vital Statistics, 1985. unpublished. variability, statistical confidence bounds for A can also be calculated. For the calcula- tion shown in Table 1, the estimated 95,.percent confidence interval on a for all smokers was 72.1 to 88.6 percent. The corresponding confidence interval for D was 27.900 to 34,300 deaths. Only 2.6 percent of the variance of the logistic transformation of a was due to sampling variability of prevalence rates. 126 Uncertainties in Attributable Risk Aggregation Bias Versus Statistical Precision Sampling variation is not the sole source of uncertainty in estimates of attributable risk. The computations of Table 1 entail the assumption that the relative risks rk depend only upon the specified indices of current and former smoke exposure. Thus, for former cigarette smokers in Table 1, the degree of risk after cessation of smoking is shown as depending only upon the length of cessation. Yet the magnitude of the residual risk also depends upon the extent of prior cigarette smoke exposure (Hammond 1968; Lubin et al. 1984) and the reason for stopping (Kahn 1966). Also, some persons may have quit smoking after lung cancer had been diagnosed. As Table 1 shows, women who had stopped smoking for 16 or more years at the time of enroll- ment into CPS-II had a subsequent 4-year relative risk of lung cancer equal to I .8. Within this group of long-term quitters, however, those women who had previously smoked 21 or more cigarettes daily had an estimated relative risk of 4.0 (Garfinkel and Stellman 1988). Likewise, for current smokers in Table 1, the degree of lung cancer risk is shown as depending only upon the current number of cigarettes smoked per day. Yet the risk depends critically upon the lifetime dosage of cigarette smoking, especially the dura- tion of cigarette use and the age of initiation of regular smoking (Brown and Kessler 1988; Doll and Peto 1978, 1981; Peto 1986; US DHHS 1982). While the relative risk r was 22.3 for all women currently smoking 31 or more cigarettes daily (Table l), it was 18.9 for heavy smokers of 18 to 30 years' duration and 38.8 for heavy smokers of more than 40 years (Garfinkel and Stellman 1988). A more detailed, multidimensional breakdown of exposure levels may minimize er- rors of classification, but such disaggregation also increases the sampling variability of the estimates. Conversely, increased aggregation of exposure levels will reduce sam- pling variability. Thus, if relative risk were assumed to depend only upon present smok- ing status (current versus former), then the estimated attributable risk for female lung cancer deaths in 1985 would be 80 percent, with a confidence range of 77 to 83 per- cent. The confidence range of attributable deaths A would be narrowed to 29,700 to 32,000. Age-Standardization The relative risks in Table 1 were estimated as a ratio of age-adjusted death rates, where the age adjustment was performed by direct standardization to the age distribu- tion of nonsmokers' person-years at risk. In principle, if the relative risk is in fact age independent, then the estimate of relative risk in large samples should not be very sen- sitive to the choice of the standard population (Anderson et al. 1980). In practice, however, the estimates can depend strongly upon the standard population. For the il- lustrative calculation in Table 1, the use of the entire population of CPS-II woman-years at risk (rather than nonsmokers only) resulted in an attributable risk for lung cancer of 79 percent, with a confidence range of 75 to 82 percent (see Table 11). 127 Potential Biases in Applying the Results of Prospective Studies to the General Population Subjects enrolled in the CPS-II prospective study constituted over 1.5 percent of a11 American adults age 45 and over (Stellman, Boffetta, Garfinkel 1988). Still, they dif- fered from the U.S. population in a number of ways (Garfinkel 198.5; Stellman and Gar- finkel 1986). CPS-II entrants were more highly educated. The black and Hispanic populations were underrepresented, though less so than in CPS-I (Garfinkel 1985). As in CPS-I, institutionalized and seriously ill persons, as well as illiterate people who could not complete a questionnaire, were excluded (Lew and Garfinkel 1984). In both CPS-I and CPS-II, the overall mortality rates of the enrollees fell substantially below those of the general US. population (Hammond 1969; Lew and Garfinkel 1988). These considerations do not by themselves invalidate the use of CPS-II to estimate smoking-attributable risks for the entire American population. The critical assumption in Table 1 above is whether the estimated relative risks -not the absolute death rates &--are representative of the general population. For CHD and for all-cause mortality, CPS-I subjects who were reportedly well at the time of enrollment showed higher estimated relative risks of cigarette smoking than those subjects who said they were sick or who gave a recent history of cancer, heart disease, or stroke (Hammond and Garfinkel 1969; Lew and Garfinkel 1988). A similar elevation of relative risk in well subjects has been found for lung cancer in CPS-II (Gar- finkel and Stellman 1988). Since initially well persons had lower disease rates, the proportional effect of cigarette smoking appeared to be larger. While CPS-I and CPS- II excluded seriously ill and institutionalized persons, the magnitude of the resulting bias is unclear. In the 1980 U.S. Census, about 1.5 percent of the U.S. adult popula- tion was institutionalized. Among persons aged 65 years and over, the proportion was 5.3 percent (U.S. Bureau of the Census 1986). Cigarette smoking has been found to act synergistically with certain workplace ex- posures (such as asbestos and ionizing radiation) in the development of lung cancer (US DHHS 1985; Saracci 1987; National Research Council 1988). Such interactions may also be present in the etiology of nonneoplastic lung disease. Alcohol and tobacco likewise interact synergistically in the etiology of oral and esophageal cancer (US DHEW 1979). Moreover, cigarette smoking has been found to interact synergistical- ly with elevated serum cholesterol and elevated blood pressure in enhancing the risk of CHD (US DHHS 1983). Persons of lower socioeconomic status (SES) may be more likely to receive such workplace exposures, to consume alcohol heavily, or to have un- favorable CHD risk factors. However, if the effects of cigarette smoking are multi- plicative, then exclusion of such persons from CPS-I and CPS-II would not bias the es- timated relative risks of disease due to cigarette smoking. Conversely, if the effects of cigarette smoking are purely additive, rather than synergistic, then the exclusion of per- sons with elevated baseline disease rates would bias upward the estimated relative risks of disease due to smoking. The estimated relative risks in Table 1 are specific to women and have been stand- ardized for age. Standardization for other stratifying or confounding variables was not performed. In principle, failure to control for such variables could bias upward or 128 downward the estimated relative risks due to cigarette use. As discussed in Chapter 2, numerous attempts to control statistically for confounding and stratifying variables have not materially altered the estimated relative risks for cigarette-related diseases. In the illustrative computation of Table 1, no distinction among the races has been drawn. For both sexes, the prevalence of current cigarette use is higher for blacks than for whites. Conversely, smaller fractions of black men and women are former cigarette smokers (US DHHS 1988b). Black persons were underrepresented in CPS-II, con- stituting only 4 percent of entrants (Stellman and Garfinkel 1986). Hence, the relative risks reported in Table 1 may not be accurate for black women. Among the 38,687 adult female lung cancer deaths in 1985, a total of v92 (8.8 percent) occurred in black women. Hypothetically, if the attributable risks a among black women had been only half those of whites, then the smoking-attributable lung cancer deaths in Table 1 would be reduced from 3 1,600 to 30,300. In prospective cohort studies, mortality rates tend to be reduced in the initial year or. two of followup. This phenomenon of lower initial mortality results from a tendency to exclude persons who are sick at the outset of the study. In particular, the relative risks in Table 1 were derived from the 4-year followup (1982-86) of CPS-II subjects. Accordingly, it is possible that the planned 6-year followup of CPS-II (1982-88) will reveal somewhat lower relative risks than those reported for the first 4 years. Conversely, measurements of exposure and other personal characteristics, typically obtained at the start of a prospective study, become less accurate as the duration of fol- lowup increases. The relative risks reported in Table 1, for example, have been clas- sified according to the subjects' cigarette smoking practices upon enrollment in 1982. If many women who were current smokers in 1982 had in fact quit smoking by 1986, then the reported relative risks for "current" smokers are actually those of a mixture of current and former smokers. In the analysis reported below, the 4-year followup of CPS-II is to be compared with the 6-year followup of CPS-I. Such a comparison needs to be interpreted in light of potential biases arising from short- and long-duration followup in prospective studies. Uncertainties in Exposure Potential errors in estimated exposure rates pk are a further source of uncertainty in the computation of attributable risk a. In the illustrative calculation of Table 1, such exposure rates were derived from the 1985 NHIS, a large-scale, stratified, face-to-face household interview survey of the noninstitutionalized civilian population of the United States. Among the possible errors in NHIS estimates are: underreporting or misreport- ing of current cigarette use; inaccurate recall of past cigarette smoking; nonresponse biases due to exclusion of some persons not available for interview; and underrepresen- tation of certain population segments. These sources of uncertainty are discussed in Chapter 5. On the whole, NHIS-derived estimates of population smoking rates have been consistent with other face-to-face interview surveys(CDC 1987a). 129 Errors in the Classification of Causes of Death The estimation of attributable deaths A requires information on total deaths D. For the computation in Table 1, the latter quantity was defined as deaths in 1985 whose un- derlying cause was primary lung cancer (International Classification of Diseases, Ninth Revision [ICD-91, Code 162). Deaths from the larger class of Respiratory Cancers (ICD-9 Codes 162-165) were not used because they include pleural mesotheliomas and secondary lung cancers. Still, the use of ICD-9 Code 162 alone may not eliminate all errors of death certification. In a review of over 1,300 thoracic cancer deaths in Min- nesota between 1979 and 1981, Lilienfeld and Gunderson (1986) identified four cases of pleural malignant mesothelioma that had been classified as Code 162.9. Moreover, it is at least arguable that physicians in recent years have been reluctant to diagnose primary lung cancer in the absence of a history of cigarette smoking (McFarlane et al. 1986). While errors in disease classification and death certification of lung cancer in 1985 may be relatively minor, the same cannot be said with assurance about other diseases caused by cigarette use. Thus, deaths certified as being caused by CHD (ICD-9 Codes 41w14) may not adequately reflect the lethal consequences of cigarette use on the cardiovascular system. Many deaths from Hypertensive Diseases (Codes 401404, in- cluding Hypertensive Heart Disease, 402, and Hypertensive Disease, 404) may have been aggravated by cigarette use. Similarly, deaths certified as being caused by COPD (ICD-9 Codes 49w92 and 496) may incompletely reflect the numbers of deaths from nonneoplastic respiratory disease due to smoking. Many cases of Influenza and Pneumonia (ICD-9 Codes 480-487) may not have been lethal but for the coexistence of cigarette-induced lung damage. The major prospective studies of cigarette smoking and mortality that were initiated in the 1950s relied upon the International Classification of Diseases, Seventh Revision (ICD-7) (Hammond 1966; Dom 1959; Kahn 1966; Rogot 1974; Rogot and Murray 1980; Doll and Hill 1956, 1964, 1966; Doll et al. 1980; Doll and Peto 1976). Coding conventions have changed considerably since ICD-7 was adopted in 1955 (Klebba 1975, 1982; Klebba and Scott 1980). While ICD-7 Code 162 was reserved for lung cancer that was "specified as primary," a separate code 163 was allocated to lung can- cers "not specified as primary or secondary." In practice, however, epidemiologists and vital statisticians recognized that the great fraction of lung cancer deaths certified under ICD-7 Code 163 were primary and that deaths certified under the two codes were in fact indistinguishable. Accordingly, it was standard procedure to report combined deaths for Codes 162 and 163-a practice adhered to in the analysis below. Still, the use of the combined category 162-163 in ICD-7 may have introduced greater diagnos- tic uncertainty than the current use of Code 162 in ICD-9. Previous Estimates of Attributable Risk from Cigarette Smoking Many authors have estimated the number or proportion of deaths attributable to cigarette use, either from a single cause, a group of causes, or all causes (Ravenholt 1964, 1984; Rice et al. 1986; McIntosh 1984; Whyte 1976; Hammond and Seidman 130 1980; Doll and Peto 1981; Gatfinkel 1980a; U.S. Office of Technology Assessment (US OTA) 1985; Schultz 1986; Goldbaum et al. 1987; CDC 1987b). Doll and Peto (1981) estimated 83,000 smoking-attributable deaths from lung cancer in 1978. Rice and colleagues (1986, Table 5) estimated 270,000 smoking-attributable deaths among U.S. adults in 1980. including 86,000 from CHD, 75.000 from lung cancer, and 14.000 from "emphysema, chronic bronchitis." The Centers for Disease Control (1987b) es- timated 3 15,000 smoking-attributable deaths for 1984, including 77,000 from CHD, 93,000 from lung cancer, and 5 1,000 from "chronic bronchitis, emphysema" combined with "chronic airways obstruction." These studies differ with respect to specific causes of disease, the time period under consideration, the populations at risk, the sources of epidemiologic data, and the specific methodology for estimation of risk. Thus, some researchers have directly applied Levin's measure of attributable risk, as defined in equations (1) and (3) (Rice et al. 1986; McIntosh 1984; CDC 1987b; Goldbaum et al. 1987; Whyte 1976). In doing so, they assumed that estimates of relative risk r, derived from particular epidemiologic studies, could be extrapolated to the population under consideration. By contrast, Ham- mond and Seidman (1980) and Garfinkel(l980a) computed attributable risks directly for the CPS-I study population. In an analysis of avoidable deaths from cancer, Doll and Peto (198 1) employed a dif- ferent model. Let N denote the size of the population at risk, while D denotes the total number of deaths from a specific cause. If do denotes the cause-specific death rate among unexposed persons, then D-&N is an estimate of the number of deaths at- tributable to the exposure. To estimate attributable cancer risks for the United States in 1978, Doll and Peto ( 198 1) then assumed that the age- and sex-specific cancer mor- tality rates for nonsmokers do observed in CPS-I during 1959-72 could be applied to nonsmokers in the general population in 1978. In support of such an assumption, they note that for men, nonsmokers' cancer rates in other prospective studies (Kahn 1966; Doll and Peto 1976) closely matched those observed in CPS-I (Doll and Peto 1981). Moreover, CPS-I lung cancer rates of nonsmoking women were similar to those of U.S. women in 1950, before their lung cancer rates began to increase. Doll and Peto's method was employed by OTA (1985) to estimate attributable deaths from CHD (US OTA 1985). For cancer, nonsmoker death rates in CPS-I may well ap- proximate do for the U.S. population. But the same conclusion does not appear to be warranted for CHD (Sterling and Weinkam 1987). In fact, the use of CPS-I nonsmoker death rates yielded an estimate of 142,000 smoking-attributable deaths from CHD in 1982. By contrast, application of the Levin method gave an estimate of 9 1,000 deaths (US OTA 1985). Doll and Peto (198 1) rejected the application of relative risks derived from CPS-I to the U.S. population in 1978. Their central concern was that such relative risks had in- creased in the two decades since the start of CPS-I in 1959. Among smokers aged 60 years or more in 1965, a much smaller fraction had smoked regularly during early life. For older women smokers, in particular, only one in eight had begun to smoke regular- ly as a teenager. This proportion increased markedly in subsequent decades (Chapter 5). In view of the importance of quantity and duration of smoking in determining lung cancer risk-and especially in view of the critical role of early-life smoking in the etiol- 131 ogy of smoking-inducedcancers (Peto 1986)-it was highly likely that the relative risks for smoking-induced cancers would have increased since the early 1960s. (See also Doll et al. 1980.) Accordingly, there may be serious biases in the application of relative risks from 1960s prospective epidemiologic studies to 1980s populations. Such potential biases constitute the most serious criticism of prior studies of smoking-attributable deaths. Updated epidemiologic evidence for the 1980s is needed to address this criticism. Populations At Risk: 1965 and 1985 Table 2 and Figures 1 through 5 describe the populations at risk in 1965 and 1985. While Table 2 reports the percentages of smokers, the figures show the absolute num- bers of U.S. resident adults in each smoking category for each year. Children and young adults under age 18, who may also suffer adverse effects from cigarette use, are ex- cluded from Table 2 and the figures. In both 1965 and 1985, respondents LO the NHIS were asked, "Have you smoked at least 100 cigarettes in your entire life. 3" Those who answered affirmatively were then asked how much they smoked currently or, if they were not current smokers, when they TABLE 2.-Prevalence of cigarette smoking, persons aged 18 years or more, United States, 1965 and 1985 196.5" 19XSh I%) C%) M&S Current smokers' Former smokers Never smoked regularlyd s3.4 32.7 20.X 79.1 2.5.x 32.x Females Current smokers' Former xmokerr Never smoked regularly" 34. I `7.5 x.1 17.1 57.8 55.4 NOTE: Pre\alence e\t~mate\ for I')63 and IYXS hate been directly itandard~red tu the a@e d~rtribulion\ t,f~he U.S. resident populattons m each year. re\pect~vely (U.S. Bureau of the Csnws lY74. 19861. "Based upon 52,X73 self-reqxmw\ IO the C~garerte Smokmg Supplement 10 the 1965 National Health Interv~eu Survey. Standard erron 0.3 IO 0.4 percent for males. 0. I IO 0.2 percent for fem;tle\. Incluwn of 33.422 addltwnal proxy respon%\ resulred in the following e\tlmales: male cnrrrnt rmokers. 5 I .Y percent: male former smoker\. IY 0 percent: female current smokers, 33 6 percent: and female former smokers. 7 7 percent. hBased upon 32.X59 self-re\ponwr 10 the C~garelte Smohing Supplement TV the 19x5 &\ational Health Inlenieu Survey. Standard ernn 0.4 percent form&\. 0.3 percent for female\. `In 1965. current smokers Included all rapondents who reported a current number \moked per day. mcludmg "1~ than I per day." In 1985. current smoker\ included all rebpondent$ who .mswered affirmatively to the queQmn "Do you bmoke nou'? `in both 1965 and 19X5. the caregor) never \mokrd regularly" mcluded wo group\ of rerpondents: ( II thaw who answered negatively to the quntion "Have you ever s.moked at lea\1 100 cigarette\ m yaur hfe'?". and (2) thaw who answered affirmatively but denled ever vnokmg c!garettes regularly. In 196.5 and 1985. reqwclnely. group I accounled for 99 percent and 97 percent of all respondent\ in thv category "never smoked regularly." 132 last smoked regularly. While the NHIS for 1965 permitted proxy respondents, the es- timates in both years have been derived from self-respondents only (see Note b ofTable 2). Table 2 shows the percentage distribution among adult men and women in three categories: current smokers. former smokers, and those who never smoked regularly. Between 1965 and 1985. the proportions of current smokers declined and the propor- tions of former smokers increased. The most marked change was the decline in the prevalence of current cigarette use among adult men. In Figure I, the responses have been further divided into four categories: current smokers of fewer than 25 cigarettes daily: current smokers of 25 or more cigarettes daily: former smokers who quit within the last 5 years: and former smokers who stopped for more than 5 years. The weighted proportions in each category, tabulated by age and sex. were then multiplied by the corresponding estimates of the U.S. resident population (U.S. Bureau of the Census 1974. 1986). In 1965. there were an estimated 53.7 million adult current cigarette smokers (stand- ard error, 0.2 million), which represented about 43 percent of all U.S. residents aged I8 years or more. By 1985, there were an estimated 53.5 million adult current smokers, composing 30percent of U.S. adults. While the total numberofcurrent smokers stayed about the same. there was a shift in their distribution by sex. The number of adult male current smokers declined from 3 1.7 million (53.4 percent) in 1965 to 28.2 million (32.7 percent) in 1985, while adult female smokers increased from 22.0 million (34.1 per- cent) to 25.3 million (27.5 percent) (Figure I ). In 1965. about 28 percent of adult male smokers who were nonproxy respondents to the NHIS consumed 25 or more cigarettes per day (Figure I ). By 1985, this proportion had risen to 32 percent. For women, the proportions of heavier current smokers rose from 14 percent of nonproxy respondents in 1965 to 21 percent of smokers in 1985. The true population prevalence of smoking 25 or more cigarettes per day in 1965 is somewhat uncertain because the elimination of proxy respondents may make the sample nonrepresentative. As shown in Chapter 5. however. there was no significant change in the proportion of heavy smokers between 1974 and 1985. By contrast, the numbers of former smokers increased substantially between 1965 and 198.5. Thus, in 1965. there were about 17.6 million adult former smokers (12.4 million men and 5.2 million women). By 1985, this number had risen to 40.9 million (25.2 million men and 15.7 million women). There was an increase in the proportion of former smokers who had stopped for more than 5 years (from 49 to 63 percent of male former smokers, and from 4 1 to 57 percent of female former smokers) (Figure 1). Cigarette Smoking and Other Forms of Tobacco Use Figure 2 shows the 1965 and 1985 adult populations broken down according to the type of tobacco used. In 1965, the NHIS included questions on cigar and pipe smok- ing as well as cigarette use. The 1985 questionnaire inquired only about cigarette smok- ing. However, questions about all forms of tobacco use, including smokeless tobacco, were included on a supplement to the 1985 Current Population Survey, performed by the U.S. Bureau of the Census (see Chapter 5). 133 80 k FEMALES p 50 F5 M 40 F m - 30 & 8 20 ii 0 rj 10 2 0 I m5 1985 1985 I 985 O Current smokers, less than 25 per day El Current smokers, 25 or more per day ml Former smokers, 5 years or less ffl Former smokers, more than 5 years FIGURE l.-Populations of current and former cigarette smokers, adult men and women, United States, 1965 and 1985 SOURCE: Estimated from unpublished tabulatmns. NH& 1965 and 1985; and estimates of the resident populations of the United States by age and sex. I%5 and 1985 (US Bureau of the Census 1974. 1986). Figure 2 shows a marked change over two decades in the forms of tobacco used by men. In 1965,5.2 million men (9 percent) had a history ofever smoking pipes or cigars, but not cigarettes. In 1985, the number using noncigarette tobacco dropped to 2.7 mil- lion or 3 percent of the men. In 1965, 29 million men had a history of ever smoking cigarettes and other forms of tobacco, about two-thirds of all cigarette smokers. By 1985, the number had dropped to 5.6 million, only 1 in 10 of all cigarette smokers. Older Cohorts of Cigarette Smokers Figures 3 and 4 focus on persons aged 60 years and over, who suffer the highest in- cidence rates of smoking-related diseases. For 1965 and 1985, respectively, these groups of older persons were born before 1906 and before 1926. Among older men, as shown in Figure 3, the two-decade interval witnessed a 136~percent increase in the num- ber of former cigarette smokers. Among, older women, the number of current smokers 134 -ES FEMLES El El _lll ta Never smoked regularly Noncigarette tobacco only Cigarettes only Cigarettes and other tobacco FIGURE 2.-Populations of adult men and women classified by history of tobac- co use, United States, 1965 and 1985 SOURCE: Estimated from unpublished tabulations. NHlSs 1965 and 1985; unpublished tabulations. CPS 1985; and es- timates of the resident populations of the United States by age and sex. 1965 and 1985 (US Bureau of the Census 1974.1986). doubled, while the number of former smokers increased sixfold. Between 1965 and 1985, the population of older women with a history of regular cigarette use, past or present, increased over threefold. The NHISs for 1965 and 1985 did not ask about the age of initiation of cigarette use. However, this information is available from other sources. For 1985, tabulations of the age of onset of regular cigarette use were made from the Current Population Survey. About 69 percent of older men with a history of cigarette use, past or present, began to smoke before age 20 (Figure 4). Among older women, the proportion was 39 percent. For 1965, three sources of information provide the age of smoking initiation among cohorts born before 1906: the NHISs of 1978-80 (Harris 1983), the Current Popula- tion Survey of 1955 (Haenszel et al. 1956). and the initial 1959 questionnaire to CPS- I (Hammond 1966, Appendix tables). For older men with a history of cigarette use, about 60 percent started smoking before age 20 (range, 56 to 62 percent). For older women smokers, about 12 percent started in their teenage years (range, 9 to 15 percent). 135 14 MALES 13 12- II - IO - Q- a- 7 t3 5 4 3 2 I 0 1985 FEMALES I3 Current smokers 69 Former smokers FIGURE 3.-Populations of current and former cigarette smokers, men and women aged 60 years or more, United States, 1965 and 1985 SOURCE: Estimated from unpublished tabulations. NH& 1965 and 1985: and estimates of the resident populations of the United States by age and sex. 1965 and 1985 (US Buxau of the Census 1974. 1986). Accordingly, the period between 1965 and 1985 saw a marked increase in the num- ber of women smokers who reached the age of 60 years (Figures 3 and 4). Moreover, the number of such women who started smoking in their teens increased by about ten- fold (Figure 4). Additional data on age of initiation are presented in Chapter 5. Overlapping Populations at Risk In 1965, a total of 7 1.3 million adults had a history of regular cigarette smoking, past or present. By 1985. this count had increased to 94.4 million. These two populations overlapped. Among the adult population at risk in 1985, about 54.8 million were born before 1948, and therefore they were also aged 18 years or more in 1965. About 95 percent of the latter group began to smoke during 1965 or earlier (Harris 1983; un- published tabulations from the Current Population Survey 1985). This means that about 5 1.8 million adults, who had ever smoked in 1985, had also been at risk in 1965. The overlap is depicted graphically in Figure 5, where the diagonal lines show the populations common to both years. Among 44.1 million adult men with a history of 136 O ? 14 MALES 13 12 11 10 Q 8 Began smoking before age 20 Began smoking age 20 or older FEMALES FIGURE 4.-Populations of men and women aged 60 years or more with a history of regular cigarette smoking, classified by age started to smoke regularly, United States, 1965 and 1985 SOURCE: Estimated from Harris (I 983); Haenszel et al. (1956); Hammond (1966); unpublished tabulations. NH& I%5 and 19X5; unpubhshed tabulations. CPS 1985: and estimates of the resident populations of the United States by age and sex, 1965 and 19X5 (US Bureau of the Census 1974, 1986). cigarette smoking in 1965, about 30.8 million survived to 1985. The vertical lines show the remaining 13.3 million men who died before 1985 (standard error, 0.4 million). Likewise, among 27.2 million adult women with a smoking history in 1965 (diagonal lines and vertical lines combined), about 6.2 million died before 1985 (vertical lines). Not all of the decedents, however, died as a consequence of their cigarette use. The horizontal lines in Figure 5 show the populations of adults at risk in 1985 who were not also at risk in 1965. The estimates are 22.6 million men and 20.0 million women. These counts do not include persons who may have taken up smoking after 1965 but died before 1985. Nor do they include smokers under age 18 in 1965 and 1985. Still, it appears that in the two-decade period following the 1964 Surgeon General's Report and the 1965 Federal Cigarette Labeling and Advertising Act, some 43 million Americans took up regular cigarette smoking, either temporarily or per- manently. About two-thirds of them began to smoke by age 18. 137 MRLES FEMRLES 80 1985 smokers born after 1947 Populations common to both 1965 and 1985 Decedents by 1985 FIGURE S.-Populations of adult men and women with a history of regular cigarette smoking, United States, 1965 and 1985 SOURCE: Estrmated from Harris (1983); unpublished tabulations. NHlSs I%5 and 1985; unpublished tabulations. CPS 1985; and esttmates of the restdent papulatmns of the United States by age and sex. I%5 and 1985 (US Bureau of the Cen- sus 1974. 1986). Changes in the Cigarette Product The 1965 and 1985 population surveys did not elicit information on the type of cigarette smoked. However, there was a decline in the average tar and nicotine yield of cigarettes, at least as measured by the U.S. Federal Trade Commission (FTC) using smoking machines under standardized conditions (Chapters 2 and 5). Data on ag- gregate cigarette sales and other population surveys (US DHEW 1979; US DHHS 1980, 1981; Chapter 5) also show that the proportion of persons smoking filter-tipped ciga- rettes increased substantially. Among entrants into CPS-II in 1982, more than 90 per- cent were filter-tipped-cigarette smokers. In this group, there was an average of 18 years of filter-tipped-cigarette smoking prior to enrollment (Stellman and Garfinkel 1986). The majority of these persons had smoked nonfilter cigarettes earlier in life. 138 It remains problematic whether such changes in cigarette manufacture and patterns of cigarette smoking have substantially reduced risks to cigarette smokers. There is considerable evidence that the actual reduction in the dangerous chemicals in cigarette smoke is much smaller than implied by the FTC machine measurements (US DHHS 1988a). While there is evidence that the long-term use of filter cigarettes and low-tar cigarettes may somewhat reduce the risk of lung cancers, there are considerably fewer data on a protective effect for other smoking-induced diseases (Alderson et al. 1985; Castelli et al. 1981; Hawthorne and Fry 1978: Kaufman et al. 1983; Lee and Garfinkel 198 1; Lubin et al. 1984; Hammond et al. 1976; Wynder and Stellman 1979; US DHHS 1981; Wilcox et al. 1988; Stellman 1986a.b). During the 1965-85 period, numerous chemical treatments and additives have been applied to cigarettes during tobacco curing and storage, sheet reconstitution, puffing, casing, and cigarette assembly. The chemicals include humectants, pesticides, flavor- ings, plasticizers, ash adhesives, and other agents. Cigarette filters, plug wraps, and tipping papers have evolved. The mix of domestic tobaccos has also changed, and oriental varieties have been added increasingly to American cigarette blends. The details of these pnlduct changes remain proprietary (US DHHS 198 1). Other Changes in the Cigarette Smoking Population The present comparison of populations at risk in 1965 and 1985 has been confined to sex, age, and history of tobacco use. Still, there may have been other changes in the characteristics of persons who smoke cigarettes. Surveys such as the NHIS have consistently shown a socioeconomic gradient in cur- rent cigarette use, as measured by education, occupation, and other characteristics (US DHEW 1979; US DHHS 1980; Novotny et al. 1988; US DHHS 1988a; Brackbill, Frazier, Shilling 1988; Chapter 5). There is some evidence that socioeconomic dif- ferentials in smoking rates have widened. The proportionate decline in adult smoking rates between 1965 and 1985 was highest for people who had graduated from college and lowest for those who had not completed high school (Chapter 5). Between 1970 and 1980, white-collar men and women showed proportionately greater declines in smoking rates than their blue-collar counterparts (US DHHS 1985). Among the factors that may influence the risks of cigarette smoking are: the coexis- tence of untreated hypertension; elevated serum cholesterol: consumption of oral con- traceptives; alcohol use; diabetes mellitus; and workplace exposure to other toxic and carcinogenic agents such as asbestos and radon daughters. With respect to these fac- tors, it needs to be determined whether the typical cigarette user of the 1980s differs from his or her counterpart of the 1960s. Cigarette smokers have higher rates of alcohol use, are more sedentary, and are less likely to wear seat belts (Schoenbom and Benson 1988; Williamson et al. 1986). It is unknown whether these relationships have strengthened or weakened over the years. There is evidence in the American population of declines in dietary cholesterol, in dietary saturated fat as a percentage of total calories, and in serum cholesterol levels (Havlik and Feinbeib 1979). The prevalence of untreated and inadequately treated hy- pertension has also declined (Havlik and Feinleib 1979). However, detailed studies of 139 the clustering of cigarette smoking with other risk factors for CHD are unavailable. It remains unclear whether the observed long-term declines in hypercholesterolemia and hypertension have been more or less pronounced in cigarette smokers than in non- smokers. There is some evidence that cigarette smoking reduces therapeutic effective- ness of new pharmacologic and invasive treatments of CHD (Deanfield et al. 1984; Galan et al. 1988). Finally, in 1965, oral contraceptives were just coming into widespread use. By 1985, oral contraceptive use was prevalent among both smokers and nonsmokers (Goldbaum et al. 1987). Those Smokers Most at Risk in 1985 Were Also Smokers in 1965 In sum, between 1965 and 1985, there have been major changes in the populations of smokers at risk for cigarette-related injury. In 1965, most men who smoked ciga- rettes had also used cigars and pipes. However, by 1985 the great majority smoked cigarettes exclusively. In 1965, about 40 percent of current smokers were women. By 1985, women numbered almost half of current smokers. Moreover, the numbers of former smokers increased substantially in both sexes- especially in men. In 1965, about one-quarter of all living men (self-respondents to NHIS, age 18 or older) with a history of regular cigarette use were former smokers. By 1985, former smokers made up almost half of all living men age I8 or older who ever smoked. Finally, the two-decade interval witnessed a substantial increase in the number of women smokers reaching the age of 66 years, with a tenfold rise in the population of older women who had begun to smoke as teenagers. These changes in the population at risk have also been observed in other, nonrandom samples of the U.S. smoking population, including a recent comparison of the 1959 entrants into CPS-I with the 1982 entrants into CPS-II (Stellman and Garfinkel 1986). The percentage of male smokers who smoked 20 or more cigarettes per day in CPS II (76 percent) was higher than in CPS-I (69 percent); the percentage of female smokers who smoked 20 or more cigarettes per day increased even more from CPS-I to CPS-II (43 percent to 6 I percent). Among the 94.4 million adults in 1985 with a history of cigarette use, about 5 1.8 mil- lion smoked cigarettes as adults before 1966. The youngest of these persons is now in his or her late thirties. This group represents the vast majority of persons who are now at risk for the fatal and nonfatal consequences of cigarette smoking. Cancer Prevention Study I and Cancer Prevention Study II CPS-I. formerly termed the American Cancer Society 25-State study, began in Oc- tober 1959 and ended in October 1972. Over 1 million men and women, representing 3 percent of the population over the age of 45 years, were recruited in I, 12 I counties (Hammond l964a,b, 1966; Garfinkel 1985). Illiterate persons, institutionalized populations, itinerant workers, and illegal aliens were not recruited. More than 97 per- cent of enrollees were white. Enrollment was by family; an eligible family had to have one member over age 45. Once a family was eligible, every family member over the age of 35 was asked to participate. As a result of family-based recruitment, more than 140 three-quarters of CPS-I subjects were married. As a consequence of the eligibility rules, the age distribution of entrants peaked at 4549 years. More than one-third of par- ticipants had at least some college education. CPS-II was instituted in September 1982. The study, conducted in all 50 States, had the same enrollment plan and organizational structure as CPS-I. Over I .2 million per- sons were enrolled. As in CPS-I. subjects were predominantly white and more edu- cated than the general population. While 2 percent of CPS-I participants were black, the proportion increased to 4 percent in CPS-II. Still, black persons were under- represented. Like CPS-I participants, CPS-II enrollees were predominantly over 40 years of age. Unlike CPS-I, the mode of their age distribution was 50 to 59 years (Gar- finkel 1985; Stellman and Garfinkel 1986). CPS-II is planned to continue through 1988. Preliminary results of the first 4 years of followup (I 982-86) are available. For these 4 years, ascertainment of the fact of death among enrollees is thought to be virtually complete. However, as of July 1988, the cause of death had not been ascertained for about 9 percent of male deaths and I3 percent of female deaths. Comparison of the 6-year followup ( 1959-65) of CPS-I and the 4-year followup of CPS-II is reported below. For computation of relative risks, cause-specific death rates for CPS-I males and females have been standardized to the age distributions of man- years and woman-years of exposure during 196569. Relative risks in CPS-II were likewise computed as the ratios of age-adjusted death rates, where standardization was performed with respect to the age distributions of man- and woman-years of exposure during 1982-86. For comparison of absolute death rates (as opposed to relative risks), the age-specific rates in both studies were standardized to the age distribution of U.S. resident white males and females in 1965. For CPS-II, absolute death rates have been corrected for underascertainment of causes of death. No such correction was made for CPS-I, where death certificate retrieval is virtually complete. No attempt has been made to correct for possible noncomparability between ICD-7 (CPS-I) and ICD-9 (CPS-II). Studies of the transition between the Seventh and Eighth Revisions of the International Classification of Diseases have shown significant non- comparability (Klebba 1975, 1982). Similar results have been reported for the transi- tion between the Eighth and Ninth Revisions (Klebba and Scott 1980). Comparison of the Seventh and Ninth Revisions, however, suggests that the combined changes have been self-cancelling (Personal communication, J. Klebba to J. Harris, June 1988). Both CPS-I and CPS-II are more representative of middle-class white Americans than the U.S. population as a whole. Still, the two cohorts were derived from virtually iden- tical sampling schemes, and analysis of the entrants has shown similar demographic characteristics (Stellman and Garfinkel 1986). These considerations enhance the validity of comparisons between the American Cancer Society studies. Nonsmokers' Death Rates Table 3 reports a comparison of the age-adjusted death rates for the three leading causes of death from cigarette smoking: CHD; chronic obstructive pulmonary disease 141 (COPD); and lung cancer. For COPD and lung cancer, in particular, there has been no discernible change in nonsmokers' death rates. The relatively small changes-less than 15 percent up or down-are all statistically insignificant. The absence of significant change in nonsmokers' lung cancer rates confirms and extends the findings of Doll and Peto ( 198 1) and Garfinkel ( 198 1). For COPD, the table presents the first information on trends in nonsmokers' death rates. It needs to be emphasized, however, that the statistical test for a change in lung can- cer or COPD rates is of relatively low power. For COPD, there are sufficient data to have detected an increase of 53 percent or more in males and an increase of 42 percent or more in females at the 0.05 level of significance. For lung cancer, increases of more than 37 and 24 percent for males and females, respectively, were detectable as statisti- cally significant. In contrast to lung cancer and COPD, Table 3 shows a very marked decline in CHD death rates in nonsmokers. Over an approximate 20-year period, nonsmokers' age- adjusted death rates dropped by 64 percent in men and 69 percent in women. The ob- served decline in nonsmokers' CHD death rates is in keeping with the CHD decline in the general population. However, the magnitude of the decline is larger in the American Cancer Society subjects. Among U.S. white males, the age-adjusted death rate from CHD (standardized to the 1965 population distribution) declined by 41 percent during 1965-85. For U.S. white females, the decline was 40 percent (NCHS 1967 and unpublished; U.S. Bureau of the Census 1974, 1986). TABLE 3.-Age-adjusted annual death rates per 100,000 for CHD, COPD, and lung cancer among males and females, aged 35 years or more, who never smoked regularly, 6-year followup (1959-65) of CPS-I compared with 4-year followup (1982-86) of CPS-II Males Females Disease CPS-I CPS-IV h CPS-I CPS-II" b CHD 74s 270 479 153 420=; 41&414c (726-775jd (256-284) (467-491) (146-159) COPD 9.5 8.7 3.0 5.6 500-502.527.1'; (7.CLl2.9) (6.5-l 1.7) (3.1-5.3) (k-7.0) 490-192, 496e Lung cancer 15.5 13.6 10.3 11.4 162-163'; 162e (12.5-19.3) (10.8-17.0) (8.9-1 I .9) (9.8-1X.3) `For both CPS-1 and CPS-II, age adjustment of rates was performed by direct standardization to the age dwnbutions of L'S, resident whtte males and females, respecttvely, in 1965 (U.S. Bureau of the Census 1974). bFor CPS-II, death rates were corrected for delayed ascenamment of causes of death. Among 4.959 known deaths during 1982-86 tn male nonsmokers. death certificates had not been recetved for 439 by June 1988. Among IO, 161 known deaths m female nonsmokers, I ,41 I had not been received. `CPS-I coding, International Classification of Diseases. Seventh Reviston. dNumbers in parentheses are 95.percent confidence tntervals. eCPS-II coding, International Classification of Dtseases. Umth Revision. SOURCE: Unpublished tabulations. Amencan Cancer Society. 142 Current Cigarette Smokers' Death Rates: Lung Cancer Figures 6 and 7, respectively, show changes in the age-specific lung cancer death rates of men and women who described themselves as regular cigarette smokers on the original questionnaire for each prospective study. The death rates, depicted in each figure on a logarithmic scale, apply to all such current smokers. No adjustment has been made for differences in the number of cigarettes smoked or duration of cigarette use. The age-incidence curves in both figures show a striking crossover effect. Among older male smokers, especially those aged 70 years or more, lung cancer death rates in CPS-II exceed those in CPS-I twofold to fourfold. By contrast, among younger male smokers, especially those less than 50 years old, CPS-II death rates are about 30 to 40 percent lower. The observed crossover phenomenon appears to be consistent with long- term changes in cigarette smoke exposure among successive cohorts. The increase in lung cancer among older male smokers reflects their increased frequency of cigarette use and increased cigarette smoking in early life. The decline in lung cancer among MALES 36-38 40-44 46-48 50-64 56-m 80-64 e&Em 70-74 76-78 m-84 RGE AT EMKUMENT 0 CPS-I A CPS-II FIGURE 6.-Age-specific death rates (log scale) for lung cancer, male current cigarette smokers aged 35-84 years; 6-year followup of CPS-I (1959- 65), compared with 4-year followup of CPS-II (198286) SOURCE: Unpubhshed tabulations. American Cancer Society. Estimates for CPS-II ax preliminary. 143 younger men may reflect their increased use of filter-tipped and low-tar cigarettes. Most currently smoking men aged 35 to 39 years in CPS-II, for example, were likely to have been lifelong filter-tipped cigarette smokers, An even more striking crossover is shown for female current cigarette smokers in Figure 7. In particular, the age of crossover comes somewhat earlier. Among women smokers aged 45 years or more, lung cancer death rates have increased fourfold to sevenfold. (There were no deaths and a small number of person-years of exposure at ages 75 or more in CPS-I.) By contrast, lung cancer death rates in the very youngest cohorts, aged 35 to 44 years, have declined by 35 to 55 percent. As in the case of men, the crossover appears to reflect differential trends in cigarette smoking among succes- sive cohorts of women. 11 35-39 40-44 45-49 50-54 55-59 60-64 65-89 70-74 75-79 80-84 AGE AT ENROLLMENT 0 CPS-I A CPS-II FIGURE 7.-Age-specific death rates (log scale) for lung cancer, female current cigarette smokers aged 35-84 years; 6-year followup of CPS-I (1959- 65), compared with 4-year followup of CPS-II (1982-86) SOURCE: Unpubhshed tabulations. American Cancer Society. Estimates For CPS-II are preliminary. Current Cigarette Smokers' Death Rates: Coronary Heart Disease Figure 8 shows the proportional decline from CPS-I to CPS-II in the age-adjusted CHD death rates of current smokers and nonsmokers. The relative declines are depicted 144 separately for men and women, and for persons younger than 65, and 65 and older. CHD death rates have declined in both cigarette smokers and nonsmokers. For the predominantly white, middle-class populations under study in CPS-I and CPS-II, the overall decline among smokers and nonsmokers was greater than observed for the U.S. white population. Still, the declines in CHD mortality rates among nonsmokers were notably greater than among current cigarette smokers. The disparity is seen at all ages, but appears somewhat greater among younger persons. In contrast to lung cancer (Figures 6 and 7). no crossover in age-incidence curves is observed. The increasing smoker-non- smoker disparity at younger ages argues against a significant salutary effect of lifelong filter-tipped cigarette use. The possibility that changes in other coronary risk factors among cigarette smokers may explain their reduced decline in CHD rates needs further investigation. 90 I MALES FEMflLES 80 70 El0 50 40 30 20 IO 0 AGED Nonsmokers Smokers MALES FEMALES FIGED FIGURE S.-Percentage decline in age-adjusted death rates for CHD; 6-year fol- lowup of CPS-I (1959-65), compared with 4-year followup of CPS- II (1982-86) SOURCE: Unpubhshed tabulations, American Cancer Scaety. Estimates for CPS-11 are preliminary. 145 0 Females, CPS-I A Females, CPS-II 0 Males, CPS-II 0 Males, CPS-I FIGURE 9.-Age-specific death rates for COPD, male and female current cigarette smokers aged 45-84 years; 6-year followup of CPS-I (1959- 65), compared with 4-year followup of CPS-II (1982-86) SOURCE: Unpublished tabulations, American Cancer Society. Esumaces for CPS-II are prelmkwy. Current Cigarette Smokers' Death Rates: Chronic Obstructive Pulmonary Disease Figure 9 gives corresponding changes in age-specific death rates for COPD. In this figure, the ages are grouped into lo-year rather than S-year age ranges as in Figures 6 and 7. For male smokers, there has been a reduction in COPD death rates for ages 45 to 74 years. For female smokers over 55 years old, there has been about a twofold to threefold increase in COPD rates. Estimated Relative Risks from CPM and CPS-II For men and women, respectively, Tables 4 and 5 depict estimated relative risks in the 6-year followup of CPS-I for all-cause mortality and for I4 specific causes of death (15 causes for women, including cervical cancer). For men in Table 4, the estimated 146 relative risks for current and former cigarette smokers are given separately. For women in Table 5, the numbers of deaths and person-years of exposure among former smokers were too small to give reliable death rates for many causes. Accordingly, in conformity with earlier reports of CPS-I mortality, the death rates for current smokers are compared with those of women with any history of regular cigarette use, past or present. For both men and women, the estimates in Tables 4 and 5 are in accord with earlier reports on CPS-I mortality (Garfinkel 1980b: Hammond 1964a,b, 1966, 1972; Ham- mond and Garfinkel 1969; Hammond and Seidman 1980). Among men, former smokers have lower mortality ratios. In both sexes, relative risks for CHD are higher at younger ages. Both sexes, but to a greater extent, men, show elevated risks of other cardiovascular diseases including stroke, hypertensive heart disease, and aortic aneurysm. In both sexes, smokers' death rates are higher for bronchitis and emphysema and for seven cancers including lung cancer. The relative risk of lung cancer among current smokers in CPS-I is about 11.3 for men and 2.7 for women. The results for CPS-II, given in Tables 6 and 7, show substantial changes in the mor- tality risk of cigarette smoking over two decades. The all-cause relative risk for men has increased from 1.8 in CPS-I to 2.3 in CPS-Il. For women, it has risen from 1.2 to 1.9. These increases in overall mortality are not an artifact of the method of age adjust- ment, because CPS-II contained proportionately fewer person-years of exposure at the youngest ages than CPS-I. As reflected in Table 6 and Table 7, the relative risks for CHD death have increased for both men and women. The relative risks for men, in particular, are consistent with those reported from recent case-control studies (Kaufman et al. 1983; Rosenberg et al. 1985) and from the followup of the Multiple Risk Factor Intervention Trial (MRFIT) cohort, as described in Chapter 2. The markedly elevated relative risks for younger women in Table 7 are consistent- with those reported in a recent case-control study (Slone et al. 1978) and in a prospective study of 120,000 female nurses (Willett et al. 1987). Such consistencies across epidemiologic studies-especially cohort and case- control studies reported during the 1980s~argue against any appreciable bias in the 4- year preliminary results of CPS-II given in Tables 6 and 7. Tables 6 and 7 show consistently increased relative risks for cerebrovascular lesions among both men and women, particularly in the younger age groups. Among women under 65 years old, the estimated relative risk of death from stroke is 4.8, with a 95- percent confidence range of 3.5 to 6.5. The observed increases in risk for current smokers are reduced in former smokers. The finding of an elevated risk of cerebrovascular disease among cigarette smokers is not new. Elevated death rates from stroke were reported in CPS-I (Hammond 1966; Hammond and Garfinkel 1969) and are reproduced in Tables 4 and 5. The 1983 Sur- geon General's Report noted the association between stroke and cigarette use; no data on the effect of smoking cessation were available (US DHHS 1983). A recent prospec- tive study of 8,000 men of Japanese origin (Abbott et al. 1986) showed an elevated risk of thromboembolic and hemorrhagic strokes among cigarette smokers. While there was no clear trend of increasing risk with higher daily smoking rates, subjects who quit smoking had reduced risks compared with continuing smokers. In the prospective study of 120,000 female nurses, Colditz et al. (1988) found a dose-response relationship be- 147 TABLE 4.-Estimated relative risks for current and former smokers of cigarettes, males aged 35 years or more, 6-year (1959-65) followup of American Cancer Society 25-State study (CPS-I) Under1 of deat i; ing cause Current smokers" Former smokers" All causes CHD, age 235 (420)` CHD, age %&Id (420) CHD, age MS (420) Hypertensive Heart Disease (44&443) Cerebrovascular Lesions, age 235 (33&334) Cerebrovascular Lesions, age 354 (330-334) Cerebrovascular Lesions. age 265 (33G334) Aottic Aneurysm. Non-Syphilttic (45 1) Ulcer, Duodenal. Gaatrtc. and Jejunal (S40-542) Influenra and Pneumonia (48048 I, 4YWY 3 ) Bronchitis and Emphysema (500-502.527. I ) Cancer, Lip. Oral Cavity. and Pharynx i 1X&148) Cancer, Esophagus ( 150) Cancer, Pancreas ( 157, Cancer. Larynx ( I6 I ) Cancer. Lung ( 162-l 63) Cancer. Kidney t I X0) Cancer. Bladder. Other Urinary Organ? t IX I I 1.80 (1.75-I.Xs)b I .38 (I .33-1.42)b 1.83 (1.7&1.91) I .42 (1.34-1.49) 2.25 (2.13-2.39) 1.56 ( I .45-l .6X) I .39 (1.30-1.48) 1.27 (1.17-1.37) I .63 I.19 ( 1.361.96) (0.94-I .5 I ) I .37 ( I .2s- I .4Y) I .79 (I .ss-2.08) I.15 ( I .02-I 30) 4.1 I (3.13-5.40) 3.06 l2.244.lX) 1.x2 ( I .JS-2.27) X.81 (6.40-12.13) b.33 (3.6&l 1.13) 3.6' t2.02k.48) 2.34 ( I x1-3.021 IO.00 (X51-XSI) I I .35 (9.1~14.15) I .x4 ( I .23-2.76) 2.90 (2.014.18) 0.96 (0.85-I .08) I .02 (0.83-I .25) 0.93 tO.8Gl.08) 2.40 ( I .73-3.34) I .49 (0.9X-2.27) I 62 (1.24-2.12) IO.20 (7.34-14.17) 2.73 ( I .36-5.49) I .28 (0.53-3.08) I .30 .(0.92-l .X4) X.60 (X7-25.74) 4.96 (3.8ti.3XJ I .79 (1.11-2.87) I .lS ( I .07-2.X7) NOTE: Bawd upon I .69?.65? man-year\ ofexpowre among male wb~ects uho never smoked regularly. 3r who smoked only ctgarettes. present or past. Relatwe n\k\, ewmated with reqxct to men who never smoked regularly. have been directlv rtandardired to the see di\tnbutmn of all man-war\ of ewosure. `Refer? to ogare& \moktng statu at e&llment IOctober 1959~iarch 194). hNumber\ tn parenthew, are 9.5.percent confidence mterval\, computed on the aywmptlon that the logarithm of relative ri\k was normally distnbuted. `All dtseate code?, refer to lntematlonal Clasvficatton of Disease\, Seventh Revwon. dWhen an age range is gtven. tt refers to the age at rnrollment in 1959. SOURCE: Unpubhshed tahulatmns. Amencan Cancer Soctety 148 TABLE S.-Estimated relative risks for current cigarette smokers and for all subjects with a history of regular cigarette smoking, females aged 35 years or more, 6-year (1959-65) followup of American Cancer Society 25State study (CPS-I) Current smoker? All causes CHD. age 235 (420)' CHD, age 35ad (420) CHD, age 265 (4201 Hypertensive Hean Disease (44&443) Cerebrovascular Lesions, age 235 (33G334) Cerebrovascular Lesions, age 354 (330-334) Cerebrovascular Lesions. age 265 (330-334) Aortic Aneurysm. Non-Syphilitic (45 1) Ulcer. Duodenal, Gastric, and Jejunal(54&542) Influenza and Pneumonia (48&481,49@493) Bronchitis and Emphysema (SW502.527.1) Cancer. Lip, Oral Cavity, and Pharynx (14&148) Cancer, Esophagus ( 150) Cancer, Pancreas ( 157) Cancer, Larynx ( 16 1) Cancer, Lung ( 162-163) Cancer,Cewx Uteri (171) Cancer, Kidney (180) Cancer, Bladder, Other Urinary Organs (181) 1.23 (1.18-1.28) 1.24 (1.2Gl.28)b 1.40 (1.29-1.51) I .38 (1.29-1.74) 1.81 (I .67-l .97) 1.74 (1.61-1.89) 1.24 (I.1 l-1.39) I .25 (1.14-1.37) 1.31 (1.04-I .66) 1.27 (1.04-1.55~ 1.19 (1.06-1.35) 1.26 (1.13-1.80) I .92 ( 1.69-2.18) 1.80 (I .59-2.03) 0.97 (0.81-1.16) 1.09 (0.95-1.26) 4.64 (3.00-7.20) 3.61 (2.465.48) 1.37 (0.81-2.31) 1.52 (0.96-2.41) 0.91 (0.59-1.41) 0.96 (0.69-I .33) 5.89 (3.97-8.76) 5.85 (4.02-8.53) 1.96 (1.14-3.39) I .89 (1.163.08) 1.94 (1.02-3.69) 2.15 ( I IFI-4.23) 1.39 (1.04-1.86) 1.38 (1.07-1.78) 3.10 (0.65-14.99) 2.69 (2.14-3.37) 2.59 (2.04-3.30) 1.10 (0.83-I .47) 1.32 (1.02-1.71) I .43 (0.89-2.3 1) 1.47 (0.97-2.23) 2.87 (I .744.74) 2.31 (1.45-3.67) NOTE: Based upon 3.325,$X39 woman-years of exposure among SUbJeCtS who never smoked regularly, or who smoked only cigarettes, present or past. Relative risks. estimated with respect to women who never smoked regularly. have been directly standardized to the age distnbution of all woman-years of exposure. aRefers to cigarette smoking status at enrollment (October 195!+March 1960). "Numbers m parentheses are 9%percent confidence intervals. computed on the assumption that the logarithm of relative risk was normally distnbuted. `All disease codes refer m International Classification of Diseases, Seventh Revision. dWhen an age range 1s given, it refers to the age at enrollment in 1959. SOURCE: Unpublished tabulauons. American Cancer Society. 149 TABLE C.-Estimated relative risks for current and former smokers of cigarettes, males aged 35 years or more, 4-year (1982-86) followup of American Cancer Society 50-State study (CPS-II) Underlying cause of death All causes Current Former smoker? smokersa 2.34 (2.262.43)b 1.58 (1.53-l.64)k CHD, age 235 (41@-414)c CHD, age 35-Gtd (410414) 1.94 (I .8&2.08) 2.81 (2.49-3.18) 1.41 (1.33-1.50) I .I5 (1.55-1.99) CHD, age M5 (41G414) Other Heart Diseasee (390-398.401AO5, 415-417.42O429) 1.62 1.29 (1.48-l .77) (1.2c1.38) 1.85 1.32 (1.63-2.10) (1.18-1.48) Cerebrovascular Lesions, 235 (43CG38) age Cenebrovascular Lesions, 35-64 (430-438) age Cerebrovascular Lesions, 265 (430-438) age Other Circulatory Diseasef(440-t48) 2.24 I .29 ( I .88-2.67) (1.1&1.51) 3.67 1.38 (2.51-5.36) (0.91-2.07) 1.94 1.27 (1.58-2.38) (1.07-I .50) 4.06 2.33 (3.08-5.35) (1.81-3.01) COPD (49CM92.496) Other Res iratory Disease"(0 I O-O 12. 480-489.893) Cancer, Lip, Oral Cavity, Pharynx (140-149) 9.65 8.75 (7.OG13.30) (6.48-l I .80) 1.99 I .56 (1.52-2.61) (1.25-I .95) 27.48 8.80 (9.9675.83) (3.15-24.59) Cancer, Esophagus ( 150) Cancer. Pancreas ( 157) Cancer, Larynx (161) Cancer, Lung (162) 7.60 5.83 (3.81-15.17) (3.02-I 1.25) 2.14 1.12 (1.62-2.82) (0.861.45) 10.48 5.24 (3.61-30.43) (1.83-14.99) 22.36 9.36 (17.77-28.13) (7.43-l 1.77) Cancer, Kidney (189) 2.95 1.95 (1.924.54) (1.31-2.90) Cancer, Bladder, Other Urinary Organs (188) 2.86 I .90 (1.85-4.44) ( I .28-2.82) NOTE: Preliminary esttmates, based upon I .49 I.791 man-years of exposure among male subjects who never smoked regularly, or who smoked only cigarettes. present or past. Relative risks, estimated with respect to men who never smoked regularly. have been directly standardized to the age distributton of all man-years of exposure. `Refers to cigarette smokrng statw at enrollment (September 1982). bNumbers in parentheses are 95-percent confidence intervals, computed on the assumption that the logarithm of relative risk was normally distributed. `All disease codes refer to lntematmnal Classdication of Diseases, Ninth Revision. dWhen an age range is given. it refers to the age at enrollment m 1982. `Includes Hypertensive Heart Dtsease (4014@4). `Includes Amtic Aneurysm, Non-Syphilitic. and General Arteriosclerosis (440-441) "Includes Influenza and Pneumonia (48M-487). SOURCE: Unpubltshed tabulations, American Cancer Society. 150