Tobacco Documents Online

816 J. E. Enstrom TABLE 3. Observed relative risk of death (RR and 95% Cl) by decade among 1954/57 current cigarette smokers (n = 63,159) compared with those who never smoked regularly (n = 42,481) among U.S. veteran males for all causes and lung cancer. Shown for comparison is expected relative risk of death based on estimated cigarette smoking prevalence for veterans in Ta- ble 1, former/never ratios in Table 2, and 1954-1959 current/never ratios by age in Appendix Table 1. Percentage changes in RRs are relative to 1954-1959 RRs. The number of deaths shown is for current and never smokers combined RR by decade of follow-up Cause of death 1954-1959 1960-1969 1970-1979 1954-1979 All causes Observed RR 1.65 1.72 1.51 1.61 95%CI 1.58-1.72 1.67-1.76 1.48-1.55 L.58-L,63 Numberofdeaths 10,111 30,420 35.068 75,599 Expected RR 1.65 t.53 1.13 1.39 Observed change 0.0% +4-2°U -8.5 % -2.4% 95%CI +1.2%,+6.7% -10.3%,-6.1% -4.2%,+L1°36 Expected change 0.0% -7.3% -31.5% -15.846 Lung cancer Observed RR 10Q89 1157 10.65 11.I0 95% Cl 7.70-15.41 9S 1-14.09 8.81-12.89 9.78-12.61 Number of deaths 558 1713 1401 3672 Expecred RR 10.89 7.45 5.64 7.41 Observedchange 0.0% +61 % -2.2% r1.9°% 95%Cl -12.7%. +29.4% -19.1`ko,+18.4% -I0.2°.L.+15.8'!5 Expected change 0.0% -31.6% -48.2% -32.04;, TAIILE4. Relative risk of death (RR and 95% Cl) from all causes by decade among 1954/57 current cigarette smokers by number of cigarettes per day compared with those who never smoked regularly among U.S. veteran males. Adjusted total is a weighted average based on the smoking distribution as of 1954-1959. The number of smokers is shown in parentheses, and number of deaths is shown for current and never-smokers combined RR by decade of follow•up Amount of smoking 1954-1959 1960-1969 1970-1979 1954-1979 Tocal (n = 63,159) Adjusted RR L65 1.73 1.55 1.63 95%CI 158-1.72 1-63-1-7i 1.52-1.59 1.58-1.65 Number of deaths 10,111 3Q,4:0 35,068 73,599 1 i d = 2 814 garette/ ay (n c . ) RR 1.11 1.15 1.12 1.13 95% Cl (197-1.28 LC6-1.=4 L05-L.20 L.08-1.18 NumberofJeaths 3265 1J,163 14,863 _'3,'_96 i 272) /d ( = 9 1 9 gare[tes c ay n , - RR L23 1?9 1.19 1.23 95%CI 1-14-1.33 1-21-1.35 1 14-1 24 1.'0-1?7 Numberof deaths 3850 L' .03'_ 17,035 32.917 = 29 10 20 /d 561) ay (n - cigarettes , RR 1.63 1.66 1.51 1.59 95%CI 1.55-1.71 1.63-1.73 L47-L55 L-56-1.61 Numbcr of deaths 6332 19.19i 23,882 49,404 21-39 cwarenes/day (n = 18,090) RR 1 93 2-C3 1.83 L92 95%Cl I.SI-2.02 1.97-2.1`9 1-78-1.89 L33-1.96 Numberofdeaths 333? 16.264 19,958 41,354 ?40eigarena/day (n = 3,422) RR 2.15 2.39 1.95 2.17 95%Ct 1.95-2.37 2.26-'-53 1.83-2.08 2.09-2.25 Numberofdeachs 3516 1C,354 14,962 29,332

824 APPENDIX TABLE 8. Relative risk of death (RR and 95% CI) by decade of follow-up and age when enrolled during 1971-1992 among 1971-1975 current cigarette smokers com- pared never-smokers based on proportional haards model of NHEFS males and females for all causes of death Age when Number of RR by decade of follow-up enrolled subjects 1971-1982 1983-1992 1971-1992 Males 25-74 2131 2.L5 (1.69-2.74) (330[ 2.09 ( 1.68-2.59) [3891 2.08 (1.77-2.44) [719] 25-34 551 5.94 (0.76-46.46) [111 4.43 (1.0t)-19.54) [161 4.96 ( L77-2.44) [27[ 35-44 383 2.11 (0.70-6.36) [191 2.13 (0.92-4.94) [32] 2.08 (1.08-4. 12) 151) 45-54 499 3.93 (L79-S.62) [64[ 2.11 (117-3.51) [881 2.57 (1.68-3.93) [l52] 55-64 373 2-05 (1.26-3.34) [831 1.89 (1.31-2.72) [1321 1,88 (1.40-2.52) [215] 65-74 321 1.90 (1.37-2.63) [153] 2.25 (1.56-3.24) [1211 2.00 (1.5 7-2.55) [274] Females 25-74 3181 1.68 (1.29-2.19) [2711 1.94 ( L.59-2.38) [4691 1.83 (156-2.15) [740] 25-34 736 2 .19 (0.40-1 L.97) [61 1.91 (0.62-5.84) (13] 2.02 (0.80-5.15) (191 35-44 572 0.76 (0.29-1.95) [18[ 1.96 (0.90-f.29) [27] 1.34 (0.75-2.41) (45] 45-54 757 2.07 (1.10-3.93) [391 2.52 (1.61-3.96) [80J 2.41 (1.67-3.48) [119] 5i-64 574 2.06 (1.17-3.63) [491 2.23 ( 159-3.12) [14I1 2.15 (L61-L86) [190] 65-74 542 1 71 (1.16-2.5t) [159/ 1 40 (0.94-2.08) [208] 1,54 (L17-2.03) [3671 APPENDIX TABLE 9. Relative risk of death (RR and 95% CI) by decade among 1971- 1975 current cigarette smokers (1329 males and 1193 females) compared with never smokers (802 males and 1988 females) based for all causes and lung cancer for ages at entry RR by decade of follow-up 1971-1982 1983-1992 1971-1992 All causes Males(n=2131) RR 2-15 Z.09 2-08 95% C[ L.69-2. 74 L.68-2.59 1.77-2.44 Number of deaths 330 389 719 Females (n = 3181) RR 1-68 1-94 1-83 95% Cl I. 2 9-2 .19 1. 5 9-2 .3 8 1. 5 6-2 .15 Number of deaths 271 469 740 Both m-.des and fem:des (n = 531?) RR 1.91 2.01 L.95 95% Cl 1.61-2.25 Lt4-2.33 1.75-2.18 Numbernf Jeaths fi0l 858 1459 Lung cancer Males RR 2067 . 39.39 28.71 9511% Cl L78-153.78 5-4L-294.24 6.98-118.16 Number of deaths 24 35 59 Females RR 5.78 5.06 5.20 95% Cl 1.36-24.60 1.99-12.84 2.58-11.35 Number of dcaths 9 21 30 Both males and females Lung cancer 10.63 10.07 10.01 95% CI 3.59-31.47 4.75-21.32 5.41-18.53 Number of dearhs 33 56 89 3. E. Enstrom

818 cohort of 3168 males and 3737 females on whom smoking data were collected during t971-1975. We created one master fite for this cohort of NHEFS subjects by linking to- gether 1971-1975 interview data with follow-up interview data from 1982-1984, 1987, and 1992, and follow-up death data through 1992 [23]. Follow-up through 1992 has been completed on 97% of the men and on 95% of the women, and there were 1109 male deaths and 838 female deaths. Follow-up of some individuals has been done through July 1993 and a few deaths occurred in 1993 and are included in the totals shown here. These NHEFS data are on publically available data files maintained by NCHS [231. To determine the impact of smoking cessation, statistical analysis using the Cox proportional ha_ards model has eval- uated the relative risk (RR and 95% Cl) from death as a function of initial smoking status and time period of follow- up for the NHEFS cohort in a manner similar to that used for the U.S. veterans cohort [l4]. The follow-up period is from time of entry into the study ( 197 1-f 975) uncil death, withdrawal (date last known alive) or end of follow-up (1992-1993), a median of 19 years. The sample weights as- sociated with the oversampling of certain NHEFS subjects have noc been used in our analysis. Our results are limited to deaths from lung cancer (ICD9 16'_) and all causes. Smoking was measured directly among survivors at four dif- ferenttimes:1971-L975,1982-1984,19S7,and 1992_ RESULTS Data on smoking habits of cigarette smokers in NHEFS co- hort during 1971-1975 has been compared with follow-up smoking data from 1982-1984, 1987, and 1992 in Table 6. J. E. Entr.>m Of those who smoked cigarettes in 1971-1975, only 45%o of the men and 50% of the women still smoked in 1992. Of those aged 55-74 at entry who smoked cigarettes in 1971- 1975, only about 30% of the men and 40% of the women still smoked in 1992. These data demonstrate the large de- gree of smoking cessation among this cohort, especially for those aged 55-74 at entry. These results are consistent with other national surveys that show a similar degree of cessa- tion among persons aged 55 years and older [t81. Also, the four surveys in NHEFS show that there is good agreement between smoking status reported in 1971-1975 and 1992 for never and former smokers. Of the former smokers in 1971-1975, 100% were former smokers in 1992. Of the never- smokers in 1971-1975, 95% were never-smokers in 1992. Table 7 shows the relative risk of death (RR and 95% Cl) during 1971-1982, 1983-1992, and 1971-1992 for cur- rent cigarette smokers aged 55-74 years at entry as of 1971- 1975 compared with never-smokers for dearh from all ~ causes and lung cancer. There has been no measurable change in RR from all causes for males, femaLes. or both genders over a 20-year period. For lung cancer, the RR in- creased for males and remained the same for fenrales, based on small numbers of deaths. Table 3 shows the baseline characteristics of 1971-1975 current cigarette smokers aged 55-74 years at entry as a function of 1982-1984 cigarette smoking status. Compared with the 1982-1984 respondents, (current and former smokers), the 1982-1984 nonrespondents have somewhat poorer health status and less education and other differ- - ences, but none of the differences are large. The 1982-1984 current and former smokers have fairly similar initial char- acceristics [deally, it would be useful to know the trends in TABLE 6. Cigarette smoking trends among NHEFS males and females. Percentage of cur- rent cigarette smokers among all tlre respondents to the original 197 I-1975 questionnaire are shown by age at entry, along vith percentage among the surviving respondents to the follow-up questionnaires in 1982-1984, 1987, and 1992. Number of respondents are shown in parentheses Percentage of current smoken by year Ageatentry 1971-1975 1982-1984 1987 1992 Males 25-35 100 (360) 79 6'_ 54 35-44 l00 (25l) 76 56 47 45-54 lQ0(353) 75 53 4l 55-64 100 (229) 69 44 3Z 65-74 t00 (165) 70 47 3 1 25-74 l000358) 74 56 45 Tutal re>pondents (1358) (1109) (857) (7S0) Females 25-35 100 (370) S1 66 59 35-44 100(269) 81 64 50 N 45-54 100(328) 80 58 46 O 5 5-64 100080) 69 44 32 , ~ 65-74 100(82) 70 47 31 ~ 25-74 100(1229) 77 59 50 Go Total respondents (1229) (998) (893) (820) N i cc

T 820 TABLE 8. Key 1971-1975 characteristics for NHEFS subjects who smoked cigarettes and were aged 55-74 years as of 1971-1975 by their 1982-I984 cigarette smoking status. No response includes five males and six females who stated they never smoked 1971-1975 characteristic Males Number of subjects Race (% white) Marital status (% married) Education (% ? 12 years) Height (mean in inches) Weight (mean in pounds) Health status (% fair/poor) Alcohol consumption (°a never) Recreational exercise (°o much/moderate) Other physical acriciry' (:S very active) Females Number of subjects Race (% white) Marital status (% married) Education (% ~ 12 cears) Height (mean in inches) Weight (mean in pounds) Health status (% fair/F vr) Alcohol consumption t'•, never) Recreational exercise t°o much/moderate) Other physical actiricy (°. very active) California (L9,B99 male cigarette smokers and 21,137 fe- male cigarette smokers). Over a 38-year follow-up period (1960-1997), about 93% of the 1959 male smokers stopped smoking, and yet the RR of the 1959 >mokers as a whole versus never-srnokers declined by onLy 11% for all causes (from 1.95 to 1.74) and by only 5% for lung cancer (from 12.63 to 11.96) [241. Over the 38-pear follow-up period, 1982-1984 cigarette smoking status Current Former No response Total 124 80 185 394 81 93 78 82 81 89 75 79 41 36 31 36 69 68 67 6S 164 168 160 161 30 35 42 37 17 20 24 21 62 52 49 54 40 35 32 35 L08 63 85 262 86 90 82 86 60 57 53 53 39 46 34 39 63 62 61 62 147 l5J 140 142 32 29 35 32 23 27 34 23 47 37 45 44 32 35 28 32 J. E. Ensrrom about 93% of the female smokers stopped smoking and yet the RR of the smokers versus never-smokers increased by 4% for all causes (1.47 to 1.53) and by 146% for lung can- cer (2.49 to 6.12) [241These mortality trends do not agree well with pcedictions of disease prevention due to smoking cessation that are based on the widely accepted results for self-selected former TABLE 9. Cigarette smoking prevalence and lung cancer death rates during 1966-1968 and 1986 among U.S. whites aged 45-64 years as of 1966-1968 based on NMS White males White females Subjects 1966-1968 1986 1966-1968 1986 Percencage of current cigamtte smokers 52 15 33 13 Percentage ot current cn_arrae smokers relatioe to 1966-1968 L0o 29 lol'. 39 Percentage of former smokers among those .••ho ever smoked cigaretces 31 79 19 65 Estimated lung cancer,learh rates (annual deaths per 1 \\OCO) and rntiu [number o( eample deaths shownl Total population 18.5 (7061 LL3 (2131 0-4 l50] 72.5 1251 Never smokers 18.5 127,1 70.0 [101 69 [30] 37.5 (131 Total/Never ratio 6.4 5.9 3.0 4.6

Srnoking Cesaation and Morr-,tleryTrcnds characterisrics for all subjects since 197I-1975, but such data are nor available. Based on the existing data, it is rea- sonable to assume that the mortality patterns observed are primarily related to smoking cessation, although other fac- tors may have played some role. DISCUSSION This study provides valuable new evidence regarding the long-term relationship of smoking cessation to the preven- tion of smoking-related diseases in two U.S. populations. First, we confirmed among the U.S. veterans cohort the well-established decrease in relative risk of death among self-selecred former smokers compared with never-smokers, when classified by the number of years quit as of the start of the study [4,10,11 [. ~, We then showed that there has been relatively little de- cline in relative risk over time among all veterans who smoked cigarettes at the start of the study compared with never-smokers, in spite of a large portion of the smokers having quit by 1979. In particular, there has been no de- cLine in the lung cancer death rate of the cigarette smokers relative to never-smokers, even though lung cancer is the disease most strongly linked with cigarette smoking and is the one that should be most impacted by cessation over a 819 long follow.up period. Our lung cancer trend for the U.Ss veterans cohort agrees with the trend recently published us- ing a slightly different definition of cigarette smokers [13J. These results are supported by a national sample, the NHEFS cohort, which also shows no decline in relative risk of death among cigarette smokers as a whole in spite of sub- stantial smoking cessation. The smoking patterns among the NHEFS cohort were measured four times, and the resultant cessation trend agrees fairly well with the imputed cessation trend for U.S. veterans..Both studies indicate that about 70% of current ma[e smokers aged 55-64 quit after 20 years. Obviously, the U.S. veterans study has the major weakness that smoking status was not measured after initial enrollment in 1954/57. However, several national surveys have shown that since 1955, a high degree of smoking cessation occurred in U.S. veterans as a whole, as well as in the general population. Thus, it is highly unlikely the veterans in the cohort have not also experienced a similar degree of cessation. In any case, the differences between the observed and expected trends in relative risks in Table 3 are so large that they would still be substantial even if cohort veterans had a smaller degree of cessation than that estimated. These findings are further supported by our new results among the Cancer Prevention Study (CPS) I cohort in TABLE 7. Relative risk of death (RR and 9i°6 CI) by decade among 1971-1975 current cigarette smokers (390 males and 259 females) compared with never-smokers (304 males and 857 females) based for all causes and lung cancer for ages 55-74 years at entry All causes Males (n = 694) RR 95%C1 Number of deaths . Females (n = 1 1161 RR 95 % Cl Number of death> &rzh males and females (n = 1810) RR 95`*iCI Number of dc:uh- Lun9cancer Male RR RR by decade of follow-up 1971-1982 1.9' 1.46-'-.52 '_36 L-19 1.31-2.46 '_i)9 1-56 1.i'_-2.25 444' 15.76 95% Cl 2.C6-I20.62 Numhcr uf deaths 15 Females RR 2.92 95% Cl 0.57-15.06 Numbcrofdcaths 6 &)dr males and fem.des Lung cancer 7.78 95% Cl 2.46-24.57 Number of.leaths 21 1983-1992 1971-1992 2.08 1.96 1.60.2.70 1.63-2.36 253 489 L82 1.79 1.42-2.34 1.47-2.17 349 557 1.93 1.88 1.62-2.31 L65-2.5 602 1046 30.23 2Z.20 4.00-228.58 5.31-92.92 23 33 6.08 4.74 2.06-17-79 1.94-1159 15 21 10.39 9.21 4.36-24.73 4.61-15.40 38 59 N I (D

5moking Cessation and Mortality Trends 1954-1959 RR is 1.65 (1.58-1.72) for all causes and 10.89 (7.70-15.41) for lung cancer, and these RRs are similar to the corresponding RRs for former smokers who had quit for <5 years. Comparing 1954-1959 with the entire follow-up period of 1954-1979, the RR declined by only 2% (1.65 to 1.61) for all causes of death and increased by 2% (from [0.89 to 11.10) for lung cancer. The 1954-1979 RRs repre- sent the cumulative impact of smoking cessation on this co- hort, as they cover all smokers for the entire follow-up pe- riod. The results show only a slight convergence of the RR for veterans who smoked cigarettes at enrollment compared with veterans who never smoked regularly in spite of fact that about two thirds of all the current cigarette smokers at enrollment were estimated to have quit 20 years later, based on Table 1. Table 3 also compares the observed convergence with e,rthe convergence that is expected based on the conven- tional results for former smokers. The expected RRs are based on the estimated cigarette smoking prevalence over time for veterans in Table 1, the former/never ratios in Ta- ble 2, and the 1954-1959 age-specific current/never ratios (1.65 for ages 55-64, 1.57 for ages 65-7-F, and 1.19 for ages 75-84). Sample calculations of the expected RRs for 1960- 1969 and 1970-f979 are shown in Appendix Table 1. Per- centage changes in RRs are shoN~ n relative to RRs for 1954-1959. The observed and expected changes for 1954- 1979 were -'-.4% versus - 15.8% for all causes and +1.9% 817 versus -32.0% for lung cancer. The observed RR changes do not come close to the predicted impact of cessation. When the RR over time is examined by number of ciga- rettes smoked per day in 1954/57, a pattern similar to that in Table 3 is seen in Table 4 for all causes of death and in Table 5 for lung cancer. There is no reduction in RR at any level of smoking, even for light smokers, who probably ex- perienced an even higher degree of cessation than the heavier smokers. Indeed, the adjusted RR based on the smoking distribution at entry decreased by only 1"o for all causes (1.6i for 1954-1959 versus 1.63 for 1954-1979) and increased by 5% for lung cancer (from 10.9 to 11S). 1971-92 NHANES I EPIDEMIOLOGIC FOLLOWUP STUDY METHODS The NHANES I Epidemiologic Followup Smdy (NHEFS) is a prospective cohort study based on the National Health and Nutrition Examination Survey (NHANES 1) con- ducted by the National Center for Health Statistics INCHS) on a representative sample of the civilian nonin,tirutional- i=ed population of the United States aged 25-74 %ears (21- 23j. Persons estimated to be at risk of malnutrition (chil- dren, the elderly, women of child-bearing age. and the poor) were oversampled. The subject of this report is the TABLE 5. Relative risk of death (RR and 95% CI) from lung cancer by decade among 1954/57 current cigarette smokers by number of cigarettes per day compared with those who never smoked regularly among U.S. veteran males. Adjusted total is a weighted average based on the smoking distribution as of 1954-1959 RR by decade of follow.up Amount of smoking I954-1959 1960-1969 1970-1979 1954-1979 Total ~~ Adjusted RR 10.9 11.6 11.4 11.3 95%0 7. 7-15.4 9.9-14.1 9.4-13.8 1C 1-13.1 Numherofdearhs 553 1713 1401 3672 <t dgarettc/day RR 3.15 4.3T 3.41 3.=5 95% Cl L=1-7.17 '_.91-6.56 2.20-5.Z9 '.>b-i.00 Number of deaths 41 136 140 317 1-9 cigarettes/dav RR 4.-1 3.69 3.76 . 95%Cl -.: -7.14 2.77-4.93 2.83-4.99 ?.16--1.58 Numur oi dcarhs 66 13J 198 452 10-30 eigtrettrs/da~ RR S 0: 60.-47 10.30 1..1 ; 95 % Cl 6.1 °-1'_.S0 5.54-1'-.35 8.43-12.58 ; Qi-I1.66 Number.ttdeach, _35 759 697 li2l 21-39 eigarettes/d:n RR lePo 17.27 17.49 l-.?t 95% CI 1 L: 3-24.39 14.05-21.2 2 14.26-21.45 1-•.' 1-19.92 Numberuf,lcatlu _6i 75' 616 1628 ?40 eigarettes/dac RR '-±.~3 '_4a9 20.32 ' 1.Jl 95% Cl 15.6 .-36.69 19.35-31.36 15.45-26.73 L 9.45-27.20 Number of deaths 92 272 214 i73

$rnokingCessation and Mortality Trends TABLE I. Cigarette smoking prevalence among World War I veterans and all males in the United States boon during 1590-1899 (aged 55-64 years as of 1955) Subjects 1955 1967 1976 1985 Percentage o(current cigarette smokers by year WW 1 veterans [refs. 13-161 48 32 15 9 All males [refs. 13-16) 46 30 14 9 All males [ref. 18] 49 30 15 Percentage of current cigarette smokers relative to 1955 - WW I veterans [refs. 13-161 IcYi 67 31 19 All males [refs.l3-16) 1C0 65 30 20 All males [reL 18] 1O0 61 31 Percentage of former smokers among those who ever smoked cigarettes W W l veterans [13-16] 27 49 76 84 AII males (t3-16j 25 49 76 83 pared with persons who never smoked regularly [14]. The follow-up period for the U.S. veterans cohort has been de- fined to be from January 1, 1954, through December 31, 1979, because follow-up for 1980 ma}have been somewhat incomplete [12[. Relative risks have been determined by decade (1954-1959, 1960-1969, and 1970-1979) and for the entire 26-year follow-up period (1954-1979). The anal- ysis has been restricted to veterans born during 1890-1899 (those aged 55-64 years as of January 1, 1955) in order to re[are more precisely mortality trends to imputed smoking trends described later here. The results are limited to deaths from lung cancer (ICD7 162-163), the cause of death most strongly related to smoking, and all causes. Because smoking habits were determined only at time of enrollment in 1954 or 1957 for this cohort, we have esti- ated smoking prevalence over time among \YIW I veter- ans born during 1890-1899 from four national surveys that provide data for U.S. male veterans and U.S. males by year 815 of birth: the Current Population Survey (CPS) tor 1955 [15],.1967[16(, and 1985 (171, and the 1976 National Health Interview Survey (NHIS) [18]. Similar analyses of smoking trends by birth cohort among all U.S. males havc been done recently using NHIS [19,20]. RESULTS Table I shows cigarette smoking prevalence during 1955- 1985 and degree of cessation since 1955 for all males and WW I veterans born during 1890-1899 based on national surveys. The cigarette smoking prevalence among WW 1 veterans declined from 48% in 1955 to 9% in 1985 among the 1890-1899 birth cohort. These data show that r-ererans had smoking patterns similar to those of all men. This high degree of smoking cessation is generally seen anton_ all per- sons over 5 5 years of age in national surs-eys [19,2C]. Table 2 presents data on the mortality parterns amonq those veterans who were already former cigarette smokers at the beginning of the study. The relarive risk of dea th dur- ing 1954-1959 by length of time quit as of entrV date (1954/57) is shown for all former cigarette smokers (n = 31,162) compared with those who nexer smoked regularl)(n = 42.481) among U.S. veteran males. The RR for all causes declines from 1.61 (1.50-1.72) for <5 year< to 1.05 (0.97-1.14) for 15+ years and the RR for lung cancer de- clines from 10.18 (6.83-15.18) for <5 years to 1.73 (0.93- 3.22) for 15+ years. Reduced RRs occur only amon, those former smokers who have survived for at least 5 years after stopping. These data are used here to predict the e\pected impact of smoking cessation on the smokers in this cohort. Findings similar to these were published more than 30 years ago and helped establish the relationship between smoking cessation and mortality [10]. Table 3 shows the relative risk of death by decade amon-1 1954/57 current cigarette smokers (n = 63,159) compared with those who never smoked regularly (n = 4'_.481). The TABLE 2. Relatise risk of death (RR and 95% Cl) during 1954-1959 by length of time quit as of entry date (1953l57) for all former cigarette smokers (n = 31,162) compared with those who never smoked regularly (n = 42,481) for all causes and lung cancer. The number of deaths shown is for current and never smokers combined RR by length of time quit in years Cause of death <5 5-9 10-14 ?15 Number of (ornrer snrokers 10,689 6303 4012 10,155 All c:wses of dendt RR 1.61 1-39 1.28 1-05 959;r CI 1.50-1.72 1.28-1.52 1.15-1.43 0.97-1.14 Number"f dearhs 4205 3652 3403 3803 Lung cancer RR 10.18 6.79 2.89 1.73 9575 Cl 6.83-15.18 4.20-10.96 1.38-6.02 0.93-3.22 Number of dearhs 117 67 43 45

7 Smuking Cessacian and Murtality Trends smokers. Thus, the impact of smoking cessation ori smokers as a whole appears to be substantially less than that univer- sally stated by the Surgeon General and other official agen- cies. !t appears chat the risk of death associated with ciga- rette smoking status at enrollment is more fixed than previously believed. These findings can be applied in an approximate way to smoking cessation and lung cancer trends in thee genera[ U.S. population- Table 9 shows the cigarette smoking pat- rerns and lung cancer death rate during 1966-1968 and 1986 for the birth cohort of U.S. whites aged 45-64 as of 1966, based on the National Mortality Survey (NMS) /16,251. In addition, the lung cancer death race is presented for chose in this birth cohort who never smoked cigarettes, based on a small number of sample deaths. The smoking data indicate that about 70% of the male smokers and 60% ~,,of the female smokers in this birth cohort quit from 1966 to 1986. However, the lung cancer mortality rate ratio of total population to never-smokers shows only slight convergence for males (from 6.4 to 5.9) and divergence for females (from 3.0 to 4.6). Although this analysis is admittedly crude, it suggests that smoking cessation has had little impact on the lung cancer death rate in this general population birth co- hort as a whole. The results for all causes of death could be affected by life- style changes other than cessation among smokers that might have negated the benefits of smoking cessation, but it is not clear what these changes might have been. Other than Table 8, data to address this issue are not.available. If it is eventually determined that other lifestyle changes ne- gate the impact of cessation, then this would become an important consideration when evaluating che overall im- 821 pact of cessation. It is much less likely that other lifestyle changes would have affected the results for lung cancer, as cigarette smoking is by far the strongest risk factor for this diseaset lr is possible that while the RR for long-term quitters de- clines over time, the RR for continuing smokers increases over time, thereby resulting in a relatively constant RR trend for the original cohort of smokers as akt'hole. This no- tion is supported by the unexplained increase in the RR for current smokers versus never-smokers since the 1960s [25,261. Another factor that may have attected the RR trends for all causes is the unexplained decline in age-spe- cific death rates from all causes (but not from lung cancer) that has occurred among never-smokers in the general pop- ulacion since the 1960s [25,26[. Regardless of these consid- erarions, the fact remains that there is a lack of conver- gence in RR in spite of substantial cessation. Although Ihese findings are subject to limitations, they raise an im- portanr issue about the value of smoking cessation. Addi- tional research is needed to understand fullc the impact of smoking cessation. APPENDIX The results for U.S- veterans aged 30-84 at entry are similar to the ones presented in this article for those aged 55-64 years at entry and are presented in Appendix Tables 2-7. Results for several smoking-related causes of death are also presenred. The NHEFS results for those aged 25-74 at entry are similar to the ones presented above for those aged 55-74 years at entry and are presented in Appendix Tables 8 and 9. ~ APPEVDhS TABLE I. Expected relative risk [RRe] for veterans born during 1890-1899. RRe is based on the age-specific RR for current smokers, RRc (age), RR for former smokers (RR(F<5], RR 1F5-91, RR [F10-14], and RR 1Fl">+1) in Table 2. The calculation is a weighted average of the excess risk in current and former smoking categories for each decade, assuming a(in- ear amount of cessation since 1955 based on the smoking data in Table l: 15% by 1960-1964, 30% by 1965-1969, 45% by 1970-1974, and 60% by 1975-1979. The calculation derails are shown for all causes of death and lung cancer for 1960-1969 and 1970-1979 All causes RRe ( [960-19691= I + [RRc(65-74) - 11 x (0.70 x (RRc - I/ + 0.15 x (RR(F<5) - 11 - 0.li x (RR(F5-9) - Ill %[RRc - 11 = 1+( L.57 - 11 x[0. 7J x [1.65 - 11 + 0.15 x [1.61 - 1I - 0.15 x(1.39 - ]]11[1,65 - 11 = 1.53 RRe(1970-1979) = I + (RRe( 7S - 84) - [ 1 x (0.-10 x [RRc - 11 + 0.1) X (RR(F<5) - 11 + 0.15 x IRR (Fi - 9) - ) ] + 0.15 x IRR(F10 - 14) - 1] + 0.15 x[RR(FI5+) - 1]l I(RRc - I I = I + [I.19 - 11 x[0.40 x[1.65 - 11 + 0.15 x/ 1. 61-11 ` 0.15 x11.39 - I]+ 0.15 x[1.28 - I]+ 0-15 x (1.05 - III/(1.65 - [1 = l.[3 Lung cancer RRe (196J-1969)= I + [RRc(65 - 74) - l ( x [0.70 x IRRc - [I + 0.15 x [RR(F<5) - I ] + 0.15 x [RR(F5 - 9) - l ll / (RRc - 1 ] = I + [7.96 - 1] x[0. 70 x[ 10.89 - 11 + 0.15 x(10.13 - 11 + 0.15 x(6.79 - l Il l(10.S9 - 11 = 7.43 RRe (1970-[979)= I+[RRc(75 - 84) - 1[ x[0.40 x[RRc - 11 + 0.15 x(RR(F<5) - 1] + 0,15 x IRR(F5 - 9) - 11 + 0.15 x [RR(Fl0 - 14) - ll + 0-15 x[RR(FL5+) - 111 / IRRc - I] = I+[7.96 - 11 x[0.40 x l 10.89 - l]+ 0.15 x(1 Q 15 - 1]+ 0.15 x[6. 79 - I]+ 0.15 x (2.89-11+0.15x[1.73-111/[10.89-1]=5.64

82Z ) . E. Enstrum APPENDIX TABLE 2. Relative risk of dcath (RR and 95% CE) by length of time quit as of entry date (1954/57 ) for all former cigarette smokers (n = 31,162) compared with drose who never smoked regularly (n = 42,481) among U.S. veteran males born during 1890-1899 for selected causes of deatit' Cause of death All former smokers All causes All cancer BC, eso, or ca Lungcancer Orhercancer All CVD Coronary HD Stroke Bron, emp, asthma n = 10,689 1.61 (L.50-1.72) [4205[ 1.82 (1.57-2. l Z) [805] 4.19 (2.00-8.79) [28[ 10.18 (6.83-15.18) (1171 1.20 ( L.00-1.44) [660] 1.54 (1.41-1.67) [2747] 1.50 (1.36-1.66) [19221 1.35 (1.03-1.78) (282[ 14.97 (8.07-27_76) [581 RR by length of time quit in years 5-9 n = 6303 1.39 (1.28-1.52) [3652[ 1.65 (1.36-L.99) 16941 2.01 (0.66-6.10) [18] 6.79 (4.20-10.96) (671 1.30 0 .04-1.61) [6091 1.40 (1.26-1.56) (24181 L40(L23-L59)[1699] f.05 (0.73-1.53)['-471 7.13 (3.31-15.38) (26[ 10-14 n = 4012 1.23 (L15-L43) [34031 1.36 (1.06-1.74) [632] 0-78 (0.10-5.96) [151 2.59 (1.38-6.02) [43[ 1.27 (0-97-1.66) [574] 1.30 ( L.14-L49) [2255] 1.42 (L.22-1.65) [1601] 0.92 (0.57-1.50) [233] 4.29 (1.53-12.04) [18[ ?15 n = 10, 153 1.05 (0.97-1.14) [38031 1.21 (1.02-L44) [723[ 0.88 (0.Z5-3.07) [171 L73 (0.93-3.21) (48] 1.19 (0.99-L.43) (6581 1.00(0.91-L11)[24941 0.98 (0.87-L.10) (17451 1.10 (0.82-L.4 7 ) [272] 3.15 ( L.3S-7.20) [23] -AII cancer (ICD7 140.Z07); lung cancer ( ICD7 162-163); buccal casity, esophayvs. and orher respimtory cancer (1CD7 140-150, 160-161); all cardlovas- cular disemes (CVD) (ICD7 330-334, 4b3-468), as vell as rmmnary hean dlscase (CHD) (ICD7 420) and stroke (ICD7 330-334): and bronchltfs, emphy- sema and asrhmz (ICDi 241. 500-502, 527.1). r At ci; all ( APPE.WIXTABLE3. Relative risk of deatlr (RR and 95% C[) by decade among all 1954/57 current cigarette smokers (n = 63,159) compared with those who never smoked regularly (n = 42,481) among U.S. veteran males born during 1890-1899 for selected causes of death RR by decade of follow-up Cause of death 1954-1959 1960-1969 1970-1979 1954-1979 A[i causes 1.65 (L.58-L.72) (10,1111 1_72 (L67-L76) (30,4201 1.51 (1.48-L55) (35,0681 L61 (1.58-L63) [75,599[ AII cancer 1.99 (1.80-2.19) (21361 2.15 (2.03-2.28) [60731 1.87 (1.78-1.97) [6313I 2-00 ( L.93-2-07 ) (14.5221 BC, eso, or ca 5.30 (3.03-9.26) [1191 8.81 (5.80-13.37) [3021 3.96 (2.95-5.32) [2S4[ 5.47 (4.4(L6.S1) [705) Lung cancer 10.89 (7.70-15.41) (5581 11.57, (9.51-1409) [17131 10.65 (8-31-12-89) [14011 1 L10 (9]8-1 2.61) [36721 Orher cancer 1.31 (Lf7-1.45) (14591 1.36 (1.28-1.45) [4058] 1.30 (1.23-1.38) (4628[ 1.33 (1.27-1.38) [10,145[ All CVD 1.58 (1.50-1.66) [6474] 1.52 (148-1.57) [1$313] 1.34 (L30-1.38) [20,454) 1.44 (1.41-1.46) [45,241] Coronary HD 1.57 (1.48-1.63) [4 i i9/ 1.50 (1.45-1.56) [11,9261 1.'-9 (1.24-1-34) [11,947[ 1.41 (L.38-L.45) [28,432] Stroke 1.35 (L.15-1-60) [6191 1.32 (1?2-L.43) (26611 1.'_'- ( L 1 4-1 29) (41691 1.26 ((.21-1.32) [7+491 Bron, emp, asthma 7.45 (4.21-13.18) [144{ 10.03 (7.92-IL71) [10(91 11.29 (5.76-14.i5) [8'-61 10.37 (8.79-L'-.24) [1969] APPEND[X TABLE 4. Cigarette smoking prevalence among veterans and all males in the United States aged 35-44, 45-54, and 55-64 years as of January 1, 1955 <5 percentage of current cigarette smokers 1955 1967 1976 1985 Subjects by age as of 1955 [ref. 13] [ref. 14] (ref. 16[ [ref. 151 35-44 years (horn during 1910-1919) Veterans 64 55 40 24 All males 62 53 38 23 45-54 years (horn during, 1900-19c`9) Veterans 61 49 29 15 All males 58 45 27 14 55-64 years (born during 1890-1599) Veterans 48 32 15 9 All males 46 30 14 9

. r J Clin EplJemiul Vol. 52. No. 9. pp.813-325, 1999 ~„ 089543i6/9y/$-ser iront mrmr Copyright ® 1999 Elsevier Sdence Inc. All «shrs rexn~ed. ~r4 Fll Si1895-4356(99)OOOqO-2 ~ ELSE}'IER VARIANCE AND DISSENT Presentation Smoking Cessation and Mortality Trends Among Two United States Populations James E. Enstrom SCHOOL OF PUBLiC HEALTH AND )ONfSJN COMPREHENSI V E CANCER CENTER, UNIVER5ITY OF CALIFORNIL. Los ANGELES, CALIFORNIA ABSTRACT. The long-terrn impact of smoking cessation on mortality is assessed among two U.S. popula[ions a large cohort of U.S. veterans aged 55-64 at entry and iollowed from 1954 through 1979 and the NHANES I Epidemiologic Followup Seudy (NHEFS) cohort of a national sample of U.S. adults aged 55-74 at entry and followed from 1971 through 1992. Direct and indirect survel' data indicate that 50-70% of those who were current cigarette smokers at entry had quit smoking dvnne the 19- to 26-year follow-up periods. The impact of smoking cessation on morcality among the cigarette smokers as a whole has been assessed by detemtining the time trend of the relative risk (RR) of death and 95o confidence interval (CI) for the cigarette smokers compared with never-smokers over the entlre follou-u,r penod in boch cohorts. The total death rates for the 1954'57 U.S. veteran smoker as a whole (63,159 mal<s) have concerrned only slightly toward those of never- smokers, from RR = L65 (15S-1.72 ) during 1954-19>9 w RR = 1 61 (1-58-1.63) during 1954-1979. The lung cancer death mtes for 19`>45 i smokers as a whole have nat convereed toward those of never-smokers. a'ith RR = 10.89 (7.70-15.41) during 1954-1959 and RR = I1.1: 9.78-12.61) during 1954-1979. The total death rates for the 1971-1975 NHEFS smokers as a whole (694 malee and 1116 females) have not converged tov,ard those of never-smukers. For males. RR = 1.92 (1.46-2.52) durins 1971-1952 and RR = 1.96 (1.63-2.36) during 1971- 1992; for females, RR = 1.79 (1.31-2.46) during 1971-19i'- and RR = 1.79 ( L47-2.17 ) during 1971-1992. The lung cancer death rates have diverged, based on small numbers of deaths. For males, RR = 15.76 (2.06-120.61) during 1971-1952 and Rft = 22.20 (5.31-92.92) durino i~-1-1992; for females, RR = 2.92 (0.57-15.06) during 1971-1982 and RR = 4.<4 (1.94-1L59) during 197!-t992. These trends are contrary to the substantial convergence predicted by the death rate trends among [.S. veterans who were former smokers at the beginning of follow-up. Whilc these results confirm chac those former smokers ..ho survive for ac least 5 years experience death rates that converge tort'ard those of never-smoker, they also indicate that a cohort of cigarette smokers that undergoes substantial cessation experiences a deathh mce that does not converge toward the death rate of never-smokers The fact that there has been no concer=ence for lun- cancer is quite surprising, as this is the disease most strongly linked to smoking and smoking .e=,uinn and less likely to be influenced by other lifestyle factors. Further investigation is needed for a complete un.lerstanding of the impact of smoking cessation. 7 cuN EnmEaucti 52;9-513-$25, 1999. ~cD 1999 Elsevier Science Inc. KEY WORDS. Smoking cessation, morralicy, U.S veter;n>, NHEFS. epidemiology INTRODUCTION 1t has been estimated b}- the Surgeon General and others that cigarette smoking causes about 19% of all U.S. deaths [fl. This represents more than 4CC,2C0 deaths per year, mainly front coronary heart duease, luno cancer, and ocher respirator3diseases, that in principle can be eliminated by smoking cessation. Hower'er, in spite of a substantial degree of cigarette smoking cesation and a large reduction in tar- and-nicotine levels in cigarettes durine the past 35 years in AJdress corrap..ndenee to: James E. Emtrom. Ph.D.. M.P.H., School uf Puhlic Hcalrh. Universiryof C.alllbmlz. Las Aneclcs. CA 90095-17 72. the general U.S. population (2(, the lung cancer death rate remains persistently high (3f. Indeed, among U.S. adults >ince (966, the percentage of current ciqarette smokers has declined from 43% to 25% and the percentage of former smokers has increased from 13% to 25°6, and yet the lung cancer death nte has increased by about 100%. Thus, it is important to evaluate the impact of smoking cessation from a perspccuce char may explain these trends. The ~alue of smoking cessation comes primarily from ob- servational epidemiological studies, which consistently in- dicate that the death rates of former cigarette smokers are lower than those of current cigarette smokers, and that the former smoker deach rates converge toward those of never- A ~ A T