In each study summarized in Table 1, the risk of oral cancer was lower among former smokers after the first few years of abstinence than for current smokers. After 3 to 5 years of smoking abstinence, oral cancer risk decreased by SO percent. In a study in Argentina (Iscovich et al. 1987) and in the large multicenter study conducted by the U.S. National Cancer Institute (NCI) (Blot et al. 1988). the risk of oral cancer among former smokers after 10 years of abstinence was comparable with that among never smokers. This observation hus been interpreted as an indication that the greatest effect of smoking on oral cancer risk may be in the later (postinitiation} stages of carcino- genesis (Blot et al. 1988). Although it is well known that smokeless tobacco (ST) increases the risk of oral cancer (Winnet al. 1981: US DHHS 1986) and that stopping the use of ST reduces the prevalence of premalignant tissue changes in the mouth (Gupta et al. 1986). there is litte information on the risk of oral cancer in former users of ST. Compared with current smokers. former smokers may have different alcohol drinking habits before and after smoking cessation, and thus comparisons of risk between current and former smokers may be confounded by alcohol consumption (Chapter 11). In three investigations, the effect of smoking cessation was examined and past alcohol con- sumption was controlled by multiple logistic regression (Blot et al. 1988: Kabat and Wynder 1989: Kabat. Hebert. Wynder 1989). In the three studies. estimates of relative risks for both current and former smokers were similar to those observed in studies in which alcohol was not included as an adjustment factor. The stability of the relative risk estimates for smoking with adjustment for alcohol intake suggests that alcohol does not substantially confound the relationship between oral cancer risk and cigarette smoking status and that the lower risk of former smokers cannot be explained by lower levels of alcohol consumption (Chapter 11). One study was sufficiently large to permit detailed stratified analysis of the modification of the smoking effect by alcohol consumption (Blot et al. 1988). In this study, former smokers were observed to have a lower risk than current smokers for both men and women at each of five levels of alcohol consumption. The U.S. Veterans Study (Kahn 1966) demonstrated that at each of three levels of past cigarette smoking exposure. former smokers had lower risk of oral cancer than did current smokers. Kabat. Hebert, and Wynder (1989) controlled for past cigarette exposure by multiple logistic regression and found that relative risk estimates, which were adjusted for past alcohol and cigarette consumption, did not differ from the crude estimates for former smokers (1.0 vs. 1.0 relative to never smokers). Second primary cancers of the mouth and pharynx occur commonly in persons with an initial primary cancer in the mouth, pharynx, or larynx. Several studies have addressed the incidence of second primaries of the mouth, pharynx, or larynx in relation to smoking status after diagnosis and treatment of the first primary. The findings of these studies are inconclusive. with some indicating reduced risk of a second primary after cessation (Moore 1965; Moore !971: Wynder et al. 1969: Silverman, Gorsky, Greenspan 1983) and others showing no clear benefit of cessation (Castigliano 1968: Schottenfeld, Gantt. Wynder 1974: Chapter S. see section on Multiple Primary Cancers). The results of two studies indicated that continued smoking after diagnosis of oral cancer may reduce survival, particularly in combination with alcohol consumption (Johnston and Ballantyne 1977. Stevens et al. 1983), These analyses, however. did not adjust for the more advanced stage of cancer among users of alcohol and tobacco at presentation (Johnston and Ballantyne 1977). The results of studies of oral cancer and cigarette smoking cessation indicate that tormer smokers experience a lower risk of oral cancer than current smokers and that this lower risk does not appear to be a result of confounding by alcohol or level of cigarette consumption prior to cessation. The risk of oral cancer has been shown to drop substantially within 3 to 5 years of cessation. Esophageal Cancer Smoking is a major cause of esophageal cancer (US DHHS 1982. 1989). In the United States. the proportion of esophageal cancer deaths attributable to tobacco has been estimated to be 78 percent for men and 75 percent for women (US DHHS 1989). As for cancer of the oral cavity. cigarette smoking is an independent risk factor for esophageal cancer but can also act in conjunction with alcohol to increase cancer risk. Table 2 summarizes the studies that have examined the relationship between smoking cessation and esophageal cancer risk. In these studies, the risk of esophageal cancer tor current smokers ranges from 1.7 to 6.4 times the risk among never smokers (median of approximately 5). These findings are similar to those for oral cancer as shown in Table |. The risks for smoking and esophageal cancer were similar among males and females. Three years after cessation, former smokers showed lower risks than current smokers in each study summarized in Table 2. with the exception of the Swedish prospective study (Cederlof et al. 1975) in which smoking-associated risks were considerably lower than in any other study. However. in followup of this cohort. more dramatic elevations in male mortality from esophageal cancer were observed in current smokers relative to never smokers: standardized mortality ratios were [.] for | to 7 g tobacco per day. 4.5 tor 8 to 15 ¢ tobacco per day, and 5.4 for more than 15 g of tobacco per day (Carstensen, Pershagen. Eklund 1987). For former smokers. the standardized mortality ratio was 1.3. Approximately 3 to S years after cessation. risk of esophageal cancer was reduced by approximately SO percent in the two studies providing information by duration of abstinence (Table 2). Data are very scant about the effects of cessation on the risk of esophageal cancer over long periods of abstinence. The ULS. Veterans Study showed that the risk among former smokers was lower at each of four levels of past numbers of cigarettes smoked per diy. A multivariate analysis in which lifetime alcohol consumption was included as an adjustment factor (La Vecchia. Liati et al. 1986) produced relative risks for current and former smokers that were similar to those observed in other studies. In this study, the crude relative risk for ex-smokers was nearly identical to one that was adjusted for alcohol consumption (2.7 vs, 3.0). suggesting that alcohol was nota confounder tn the estimates of the benefits of cessation. A study that was limited to nondrinkers (La Vecchia and Negri !989) also produced risk estimates for smoking that were very 152 TABLE 2.—Studies of esophageal cancer that have examined the effect of smoking cessation Risk relative to never smokers Yr Population (yr of Design Current Former since Reference data collecuion) (number of subjects) Gender smokers smokers quitting Comments Kahn US veterans Prospective Male 5.3 1.6 NP Excludes “doctor's orders” (1966) (1954-62) (248,195) quitters Cancer mortatity Cederlof et al. Sweden Prospective Male 1.7 17 NP Cancer incidence (1975) (1963-72) (27,300) Wynder and 6 US cities Case:control Male 3.6 4.8 1-3 Stellman (1969-75) (159:6,534) a) +6 (1977) Ld 7-10 1.3 iW oIs 1.0 216 (76:6,522) Female 5.3 10 1.3 at 4-6 i) 7 10 2.2 Has IS 216 Wigle, Mao, Grace Alberta, Canada Case:control Male S.1 1] NP (1980) (1971-73) (45:1,002) Rogot and Murray US veterans Prospective Male 64 24 NP Excludes “doctor's orders” C1980) (1954-69) (293.958) qpuitle rs Cancer mortality Extension of US Veterans Studs TABLE 2.—Continued Risk relative to never smokers — Yr Population (vr ot Design Current Former since Reference data collection) (number of subjects) Gender smokers smokers quitting Comments La Veechta. Liati Northern [tals Caseccontrol Miatle and 4.3 4 <5 Adjusted for SES. diet. etal. (1986) (1984 XS) (129:426) female 2.5 25 and alcohol La Veechta and Norther [tals Caseccontrol Male and 3.6" 11 NP Analysis limited to only Negri (1989) (L984 RR) C30 TRO) female nondrinkers NOTE: NP=not provided: SES =sociagcononie status ‘Computed as a weighted average from iyarette dose specie relative risks presented ae the paper, Weights are the number of controls within each stratum of smoking similar to those derived from other studies, supporting an earlier observation of elevated risk for esophageal cancer in nondrinking smokers (Tuyns 1983). This review of past research on esophageal cancer and cigarette smoking cessation indicates that former smokers experience a lower risk of esophageal cancer than do current smokers, and that this lower risk is not because of confounding by lower alcohol intake among former smokers. Pancreatic Cancer The association, noted for many years, between smoking and cancer of the pancreas is considerably weaker than that between smoking and oral or esophageal cancer (US DHHS 1982). Although the causal mechanisms underlying this association are unclear. smoking has nonetheless been regarded as a contributing factor in cancer of the pancreas (US DHHS 1982, 1989). In the United States in 1985. the proportion of pancreatic cancer deaths attributable to smoking has been estimated to be 29 percent in men and 34 percent in women (US DHHS 1989). Table 3 summarizes studies of the relationship between pancreatic cancer and smoking cessation. In these studies, current smokers had risks ranging from 1.0 to 5.4 times (median of approximately 2) the risk among never smokers. Risks for pancreatic cancer associated with smoking were similar for males and females. Former smokers generally had lower risk than current smokers for pancreatic cancer. but the available data do not characterize adequately the change in risk with duration of abstinence. The large case-control study conducted in Los Angeles, CA, (Mack et al. 1986) would suggest that risk is not substantially reduced until after 10 years of abstinence, whereas the smaller English study (Cuzick and Babiker 1989) suggests that substantial risk reduction is more immediate among women than among men; risk reduction may take as long as 20 years among men. This difference in the time course of risk after cessation according to gender has no clear biologic explanation and may be only a chance finding. The question of potential confounding by differences in cigarette smoking exposure prior to quitting was addressed in the analysis of the U.S. Veterans Study (Kahn 1966). In each of four levels of past cigarette consumption, the risk among former smokers was found to be lower than that among current smokers. In the study conducted by Falk and colleagues (1988), former smokers had a lower risk of pancreatic cancer than current smokers at each of three levels of numbers of cigarettes consumed per day and also at each of four levels of numbers of years smoked. Because alcohol can cause insult to the pancreas and has been thought to be a possible pancreatic carcinogen (Cubilla and Fitzgerald 1979), two investigators adjusted for lifetime alcohol consumption in multiple logistic regression analyses (Falk et al. 1988: Clavel et al. 1989). These analyses produced relative risk estimates similar to those derived from other studies that did not adjust for alcohol and thus suggested that alcohol consumption is not a confounding factor in the smoking--pancreatic cancer association. The results of epidemiologic investigations on pancreatic cancer and cigarette smok- ing cessation indicate that there is a weak, but consistently observed, association between smoking and pancreatic cancer and that former smokers experience a lower ISS TABLE 3.—Studies of cancer of the pancreas and smoking cessation Risk relative to never smokers Yr Population yr of Design Current Former since Reference data collection) (number of subjects) Gender smokers smokers quitting Comments Kahn US veterans Prospective Mate 1.6 1.2 NP Eacludes “doctor's orders” (1966) (FUS4 62) (248.195) quitters Cancer mortality Cederlot chal. Sweden Prospective (LYTA) (YOO 734 (27.300) Male 2.5 1.7 NP Cancer ineidence (277,700) Female L.0 3S NP Rogot and Murray US veterans Prospective Male I.8 1.2 NP Excludes “doctor's orders” (1980) (1954 O94 (293.958) quitters Cancer mortality Extension of US Veterans Study Mae Mahon etal Boston, MEA Caseccontrad (1981) (1974 79) (2 ER: 306) Male 1.3" 4 NP (1492337) Female 1.6" 3 NP Woader Hall, O US cities Case:control Polanski los 3) (1977 S14 (153:5 464) Mate a 1.7 2] ($21:2.5245) Female LL?" L4 2] Goldetal. Baltimore, MD Caseceontrol Male and [8 1.0 NP (IUSS) (1978 804 (20b 2014 female TABLE 3.—Continued Risk relative to never smokers —_ Yr Population (yr of Design Current Former since Reference data collection) (number of subjects} Gender smokers smokers quidiing Comments Mack et al. Los Angeles, CA Case:contro! Male and 2.3" 33 <5 (1986) (1976 81) (490:490) female 23 su 1.0" > 10 Norell et al. Sweden Case:control Mate and 1.0" Ll NP Data for population controls (1986) (1982-84) (98: 134) female La Vecchia et al. Northern Italy Case:contral (1987) (1983-86) (99:47 1) Male 1.6 14 NP Crude relative risk computed (St:134) Female 11 09 NP from dita presented Mills etal. California Prospective Male and S4 LA NP Cancer morkality: study (1988) (1976-83) (34.000) female Falk et al, Louisiana Case:control Mitle and Ls" ne 3 Adjusted tor diet and aleohol (1988) (1979-83) (63:12 44) female Clavel etal. Parts, France Case:control C19X9) (1982 85) (OX 161) Male Lo" 1a NP \dqusted for alcohol and (63:107) Female 15" 0.9 NP coffee TABLE 3.—Continued Risk relative to never smokers Yr Population (yr of Design Current Former since Reference dita collection) (number of subjects) Gender smokers smokers quitting Comments Cuvick and Babther England Case:control (1989) (1983 KO} (123:150) Male 21" 3.6 20 (93:129) Female 1.3" 0.8 20 Olsen etal. Minneapolis St.Paul MN Case:control Male 2.5" OR NP (POX9) (1980 &3) (212.220) ACS CPS-H United States Prospective (unpublished (1982 X6) (421,663) Male 2.0 2 NP Cancer mortality tabulations) (605,758) Female 2.7 1.6 Farrow and Davis Seattle, WA Case:control Male 3.2 1.0" NP Adjusted for age. race. and (in press) (1982 KO) (148.188) education NOTE. NP2aot provided. ACS CPS He Atierican Cancer Sociely Cancer Prevention Study tH. ge Computed as a werghted weenie from crearetiy dose specie relative risks presented in the puper © Weights are the number of controls within cach stratum: of smoking. risk of pancreatic cancer than current smokers. This diminution of risk with abstinence serves to strengthen the hypothesis that smoking is a contributing cause of pancreatic cancer. Although alcohol does not appear to be a confounder in the assessment of the benefits of smoking cessation. the possibility of confounding by other factors. such as diet or amount of prior cigarette consumption. has not been adequately studied. Bladder Cancer As with pancreatic cancer. the relationship between bladder cancer risk and smoking has been noted for many years. However, because relative risks have not been greatly elevated and because of uncertainty about the effects of unidentified confounding factors in this disease. the causality of this association has been considered less certain compared with other diseases in earlier reports of the Surgeon General (US DHHS 1982). Smoking has nonetheless been regarded as a contributing factor in bladder cancer: in 1985, it was estimated that in the United States 47 percent of bladder cancer deaths in males and 37 percent in females are attributable to smoking (US DHHS 1989), A particular problem with causal inference in smoking and bladder cancer arises because of the inconsistent finding of clear exposure—response relationships in all studies, as has been observed between cigarette smoking and respiratory cancers. However, the usual measures of exposure to tobacco smoke may not accurately index the bladder’s dose of tobacco-related carcinogens, The Intemational Agency for Research on Cancer (IARC) concluded. based on evidence available through 1985, that smoking of different forms of tobacco is causally related to cancers of the bladder and renal pelvis (IARC 1986). In addition to the studies reviewed in the 1982 Surgeon General's Report (US DHHS 1982) and in the 1986 report of IARC (1986), more recent data document a consistent association between cigarette smoking and bladder cancer. In an extended followup of a cohort of 25,000 Swedish males, mortality rates for bladder cancer were increased fourfold among ever smokers compared with never smokers (Carstensen, Pershagen, Eklund 1987). In current smokers, the risk of death from bladder cancer was approximately three times greater at all levels of consumption. The excess mortality from bladder cancer among current smokers was comparable in the American Cancer Society (ACS) Cancer Prevention Study H (CPS-ID) (Table 4). An extension of a large hospital-based case-control study, originally reported in 1977 (Wynder and Goldsmith 1977), showed similar increases in risk among male and female smokers (Augustine et al. 1988). The study included 1,316 male and 505 female cases and 3,940 male and 1.504 female controls interviewed in 9 U.S. cities between 1969 and 1984. For current smokers, odds ratios increased to approximately 3.5 for male and female smokers of 21 to 30 cigarettes per day. Odds ratios were lower among former smokers, although the risk did not decline as the duration of abstinence lengthened (Table 4). The findings of a recent population-based case-control study documented similar levels of bladder cancer risk associated with cigarette smoking (Slattery et al. 1988). Slattery and coworkers (1988) assessed cigarette smoking and bladder cancer in 332 white male cases and 686 controls in Utah. The overall crude odds ratio for current [S9 091 TABLE 4.—Studies of bladder cancer and smoking cessation Risk relative to never smokers Yr Population (yt of Design Current Former since Reference data collection) (number of subjects) Gender smokers smokers quitting Comments Kahn US veterans Prospective Male 1.9 Ls NP Excludes “doctor's orders” (1966) (1954 62) (248.195) quitters Cancer mortality Cederlof etal. Sweden Prospective Male L.& 2.1 NP Cancer incidence (1975) (1963 72) (27.300) Female 1.0 0 NP (27,700) Wynder and Stellman 6 US cittes Caseccontrol Male 2.7 2.9 1-3 (1977) (1969.75) (S41:6.534) Lo +6 L4 7-10 16 11-15 1] 216 (150:6,522) Female 2.4 3.1 |-3 LS 4-6 0 7-10 1.5 11-15 24 216 Wiele. Mao, Grace Alberta, Canada Case:control Male 2.8 2.1 NP Adjusted for cumulative (1980) (1971 73) (20421002) past dose (51:674) Female 3.5 3.1 NP Rogot and Murray US veterans Prospective Male 22 L4 NP Excludes “doctor's order” (1980) (1954 69) (293,958) quitters Cancer mortality Extension of US Veterans Study 19| TABLE 4.—Continued Risk relative to never smokers Yr Population (yr of Design Current Former since ; Reference data collection) (number of subjects) Gender smokers smokers quitting Comments Wynder and 6 US cities Case:control Male 22 2.6 1-3 Cases are from the same series Goldsmith (1977) (1969-74) (574:568) 2.9 4-6 as reported by Wynder and b5 7-9 Stellman (1977) 1.6 10-12 1.2 13-15 Il 2lea (155:154) Female 2.2 2.5 1-46 12 27 Vineis et al. Italy Case:control Male 6.0 37 3-9 (1983) (1978-81) (358:276) 3.6 10-14 2.1 215 Cartwright et al. England Case:control Male 1.6 10 6-15 (1983) (197%-&1) (932:1,402) I 16-25 0.9 226 (327:579) Female l.4 0.5 6-15 0.8 216 Morrison et al. Boston. MA Case:control (1984) (1976-77) (427:391) Male aye Ls 21 (165:142) Female 5.6" 34 zl Manchester. UK (398:490) Male 2.6" 1.8 >I (1976-78) (155:241) Female 21" 0.7 al Nagoya. Japan (224:442) Male 2.0" 10 >I (1976-78) (66: 146) Female 4.3" NP NP col TABLE 4,—Continued Risk relative to never smokers Yr Population (yr of Design Current Former since Reference data collection) (number of subjects) Gender smokers smokers quitting Comments Vinets, Esteve. aly Case:control Male Rn? 3." 3-9 Adjusted for number of Terracint (1984) (1978 83) (S12:596) 2.0" 10-14 cig/day 2.3" 215 Vineis et al. Ttaly Case control Female 2.3 1.0 23 (E9XS) (1981-83) (55:202) Jensen etal. Copenhagen, Denmark Case :contro! Male and 3.4 2.0 NP (1987) (1979-81) (388:787) female Brownson, Chang. Missouri Case:control Male 1.9 1.2 NP Adjusted for alcohol Davis (1987) (1984-86) (82322469) Hartge et al. United States Case:control Male and 2.9 2.2 1-9 (1987) (1977-78) (2.982:5,782) female 1.6 10-19 17 20-29 14" 230 Iscovich et al. Argentina Case:control Male and 7.2 4.5 2-4 (1987) (1983-85) (117:234) female 1.8 5-9 1.6 10-19 il 220 Augustine et al. DUS cities Case:control Male 2.2" 2.4 <6 (1988) (1969 84) (1.316:3,940) 2.2 7-12 21 213 (505: 1,504) Female 0.9" 7 <6 1.2 7-12 1.2 <13 c9Ol TABLE 4.—Continued Risk relative to never smokers Yr Population (yr of Design Current Former since . Reference data collection) (number of subjects) Gender smokers smokers quitting Comments Slattery et al. Utah Case:control Male a7 3.7 O.5-7 (1988) (1977-83) (332:686) 2.7 K-15 1.9 16-29 1.8 230 Claude. Frenzel- Germany Case:control Male 3.5 1.8 NP Adjusted for number of Beyme, Kunze (1988) (1977-84) (S31:531) cig/day ACS CPS-II United States Prospective (unpublished (1982-86) (421.663) Male 2.9 2.0 NP Cancer mortality tabulations) (605,758) Female 2.8 2.0 NP Burch et al. (1989) Canada Case:control (1979-82) (627:602) Male 27 1.7 NP (199:190) Female 2.6 1.2 NP NOTE: NP=not provided: ACS CPS-ll=American Cancer Society Cancer Prevention Study UH. “Computed as a weighted average from cigarette dose-specific relative risks presented in the paper. Weights are the number of controls within each stratunr of smoking. f 7 + ‘Includes current and former smokers who quit in the past 2 ye. “Crude (unadjusted) odds ratio calculated from tables presented in the paper. smoking, compared with never smoking, was 3.69 (95-percent confidence interval (C1), 2.58-5.26). However, an exposure-response relationship was not evident with reported average number of cigarettes smoked daily. The odds ratios for former smokers declined only after 8 years or more of abstinence. Table 4 summarizes findings from studies that have examined the relationship between cigarette smoking cessation and risk of bladder cancer. Of all the non- respiratory cancer sites, the relationship between bladder cancer risk and cigarette smoking cessation has been most extensively studied. In these studies. the risk among current smokers ranges trom 1.0 to 7.2 times the risk among never smokers (median of approximately 3): risks are similar among males and females. More recent studies conducted since the mid-1970s tend to show higher risks for current smokers than do the earlier studies. The higher risks in more recent studies may reflect the earlier age of starting to smoke of more recent cohorts of smokers (US DHHS 1989) or the presence of a long latency period for the smoking effect to become fully manifest after initiation in susceptible persons. Beyond the first few years of abstinence, former smokers generally have lower risks than current smokers, The study conducted in six U.S. cities (Wynder and Stellman 1977, Wynder and Goldsmith 1977) indicated an approximate 50-percent reduction in risk after 6 years of abstinence, with risk returning to that of nonsmokers among men after 15 years. A similar return to nonsmoker risk was also observed after 6 years of abstinence in an English study (Cartwright et al. 1983) and in an Argentine study after 20 years (Iscovich et al. 1987). However, results from other studies (Howe et al. 1980: Vineis, Esteve, Terracini 1984; Hartge et al. 1987: Burch et al. 1989) indicated that the reduction in risk in the first few years after cessation is followed by little subsequent additional reduction, even beyond 10 or 15 years of abstinence. These observations are in contrast to those for the other cancer sites reviewed in this Chapter. In some studies, the analyses controlled for the possible confounding effects of lower cigarette consumption among former smokers prior to cessation. The U.S. Veterans Study (Kahn 1966) showed no reduction in risk for former smokers, compared with current smokers, at levels of past cigarette consumption of | pack or less per day. There was an approximate 50-percent reduction in risk. however. for those former smokers who had previously smoked more than | pack per day. Most studies that included past cigarette smoking exposure as a covariate in multiple logistic regression analyses (Wigle. Mao, Grace 1980: Howe et al. 1980: Vineis. Esteve. Terracini 1984; Claude. Frentzel-Beyme. Kunze 1988: Slattery et al. 1988: Burch et al. 1989) showed relative risks that were similar to those observed in studies in which no such adjustment was made. A large multicenter study conducted by NCI (Hartge et al. 1987) contained sufficient numbers of subjects tor detailed subgroup analyses. Table 5 displays the findings of this study when both average cigarette dose per day and duration of smoking are cross-classified for current and former smokers. In each of these nine categories. bladder cancer risk was lower among former smokers than among current smokers, As reviewed above. the amount of evidence supporting cigarette smoking as a cause ot bladder cancer has become increasingly compelling since the 1982 Report of the Surgeon General (US DHHS 1982). which focused on cancer. Multiple studies of 164 TABLE 5.—Bladder cancer risk according to smoking dose, duration of smoking, and smoking status Duration of Risk relative to never smokers Smoking dose (cig/day} smoking (VF) Current smokers Former smokers <20 <20 1.7? 1.3 20-39 1.6 LS 240 27 19 20-39 <20) 2.2 L.4 20-39 3.8 1.8 240 31 2.5 240 <20 24 1.0 20-39 40 2 240 3.8 28 SOURCE: Hartge et al. (1987) varying design conducted throughout the world have shown statistically significant increases in risk of bladder cancer among smokers. Cigarette smoking. determined to be a contributory factor in bladder cancer in past reports of the Surgeon General (US DHHS 1982, 1989). can now be identified as causally associated with bladder cancer. The evidence adequately meets the criteria for causality established in the 1964 Report (US PHS 1964). The decline in risk of bladder cancer with cessation further supports the conclusion that cigarette smoking causes bladder cancer. This diminution in Tisk cannot be explained by confounding from lower cumulative consumption among former smokers compared with continuing smokers. Cervical Cancer Recently, an association has been noted between cancer of the uterine cervix and cigarette smoking (Williams and Horm 1977; Stellman, Austin, Wynder 1980: Lyon et al. 1983; Hellberg. Valentin, Nilsson 1983; Berggren and Sjostedt 1983. Peters et al. 1986: Brock et al. 1988: Nischan, Ebeling, Schindler 1988). However, because of the possibility of confounding by unidentified factors (in particular, a sexually transmitted etiologic agent), this association has not been identified as causal (US DHHS 1982, 1989; IARC 1986). Components of tobacco smoke can be identified in the cervical mucus of smokers (Sasson et al. 1985; Schiffman et al. 1987). These compounds have been found not only to display mutagenic activity in this environment (Holly et al. 1986), but also to have the ability to impair local immunity by reducing the populations of Langerhans’ cells within the cervical epithelium (Barton et al. 1988). The reduction in circulating levels of B-carotene caused by cigarette smoking is yet another mechanism whereby cigarettes may increase the risk of cervical cancer (Harris et al. 1986: Brock et al. 1988; Stryker et al. 1988). Thus, the association of cigarette smoking with cervical cancer is biologically plausible. 165 Table 6 summarizes findings from studies that have examined the relationship between cervical cancer risk and cigarette smoking cessation. In these studies, the risk among current smokers ranges from 1.0 to 5.0 times the risk among never smokers {median of approximately 2). Smoking-associated risks for invasive cancer and for carcinoma in situ are generally similar. After the first year of abstinence, former smokers have lower cervical cancer risk than Current smokers in most studies. Exceptions include the study conducted in Milan (La Vecchia, Franceschi et al. 1986), which showed risk reduction for invasive cancer but not for carcinoma in situ among former smokers, and the study conducted in Central America (Herrero et al. 1989) in which no association with smoking was observed at all, even for current smokers. The effect of time since stopping has not yet been well studied for cervical cancer, but observations from a large multicenter study conducted by NCI (Brinton, Schairer, Haenszel et al. 1986) suggested that risk reduction may occur fairly rapidly after cessation. One study found that smokers tended to have a poorer prognosis for survival after radiation treatment for invasive cervical cancer, but no data were presented regarding smoking cessation (Kucera et al. 1987). A major concern in studies of smoking and cervical cancer has been the potential for confounding by factors that would predispose a woman to become infected with a sexually transmitted agent that might be causally related to the disease. such as human papilloma virus (Stellman, Austin, Wynder 1980: Winkelstein et al. 1984: IARC 1986). Therefore, it is important to note that those studies that controlled for risk factors for sexually transmitted disease (Trevathan et al. 1983: Greenberg et al. 1985: Herrero et al. 1989: Slattery et al. 1989) produced relative risk estimates for current and former smokers that were quite similar to those from studies that made no such adjustments, The association of smoking and cervical cancer has been considered by some to be a result of residual confounding by inadequately measured indicators of exposure to a sexually transmitted agent. Although factors such as the number of past sexual partners are only surrogates for a hypothetical etiologic infectious agent, they are the very same social correlates of tobacco smoking that would suggest this type of confounding. Therefore, even though such factors as age at first intercourse and the number of sexual partners are imperfect indicators of infection by a possible etiologic agent, their inclusion as covariates in multivariate analyses may be sufficient to control confound- ing to some extent in the analysis of the effects of smoking on cervical cancer risk. This review of the evidence on cervical cancer and cigarette smoking cessation indicates that there is a consistently observed association between cervical cancer risk and cigarette smoking and that former smokers experience a lower risk of cervical cancer than current smokers, even after adjusting for the social correlates of smoking and risk of sexually acquired infections. This observed diminution of risk after cessatton lends support to the hypothesis that smoking is a contributing cause of cervical cancer. Based on a recent comprehensive review of epidemiologic studies providing data on smoking and cervical cancer. Winkelstein (1990) concluded that smoking is causally associated with cervical cancer. 166 L9[ TABLE 6.—Studies of cervical cancer and smoking cessation Risk relative to never smokers . . Yr Location (yr of Design — Current Former since Reference data collection) (number of subjects) smokers smokers quitting Comments Cederlof et al. Sweden Prospective 5.0 3.0 NR Cancer incidence (1975) (1963-72) (27,700) Clarke, Morgan, Toronto, Ontario Case:control 2.3 1.7 NR Invasive cancer Newman (1982) (1973-76) (178:855) Marshall et al. Buffalo, NY Case:control 1.6 0.8 NR (1983) (1957-65) (513:490) Trevathan et al. Atlanta, GA Case:control 4.2 2.1 NR Carcinoma in situ (1983) (1980-81) (99:288) Adjusted for sexual partners, birth control pills, SES Greenberg et al. England Prospective 3.0" 0.7 NR Invasive cancer incidence (1985) (1968-83) (17,032) Adjusted for age at marriage. birth control pills, SES Brinton, Schairer, 5 US cities Case:control (5 3.2 I Adjusted for sexual partners, age at first Haenszel et al. (1986) (1982-84) (480:797) hl 2-4 intercourse, SES 10 5-9 Lt 210 La Vecchia, Milan, Italy Case:control 14° 2.5 NR Carcinoma in situ Franceschi et al. (1981-84) (183:183) 1.7 0.8 NR Invasive cancer (1986) (230:230) 891 TABLE 6.—Continued Risk relative to never smokers Yr Location (yr of Design Current Former since Reference data collection) (number of subjects) smokers smokers quitting Comments Brisson etal, Quebec Case:control 3.5 19 NR Carcinoma in situ (1988) (1982-85) (247:137) Herrero et al. 4 Central American cities Case:control 1.0 1.0 NR Adjusted for sexual partners (1989} (1986-K7) (666: 1 427) Slattery etal. Utah Case:control 34 1.4 NR Adjusted for sexual partners and education (1989) (19%4—87) (266:408 } ACS CPS-H United States Prospective 2.1 1.9 NR Cancer mortality (unpublished (1982 -¥6) (605,758) tabulations) NOTE: NR=not reported: SES=socioeconomic status: ACS CPS-H=American Cancer Society C “Computed as a weighted aver he Computed as a weighted av ancer Prevention Study IL ge from cigarette dose-specific relative risks presented in the paper. Weights are the number of incide ge from cigarette dose-specitic relative risks Presented in the p: AL cases within cach stratum of smoking. aper. Weights are the number of controls within each stratum of smoking. Breast Cancer In general, prior research has shown little relation between cigarette smoking and the risk of breast cancer (Baron 1984: Rosenberg et al. 1984: Baron et al. 1986): however. in recent years, several reports have raised the possibility that there might be a weak positive association (Table 7). Because there has been considerable discussion about the possible role of smoking in breast cancer in recent literature. the relationships among cigarette smoking, smoking cessation, and breast cancer risk are reviewed. Cigarette smoking creates a set of physiologic conditions that result in various antiestrogenic effects (Baron 1984; Jensen, Christiansen, Rodbro 1985: Michnovicz et al. 1986). as well as affecting body mass (Carney and Goldberg 1984: Hofstetter et al. 1986: Chapters 9, 10, 11). The relationship between cigarette smoking and body mass is a particularly important consideration in studies of breast cancer, because body mass has a complex age-dependent association with breast cancer risk, with obesity being protective in premenopausal ages but slightly risk-enhancing later in life (Willett et al. 1985). Table 7 summarizes findings from studies that have examined the relationship between breast cancer risk and the cessation of cigarette smoking. The risk of breast cancer among current smokers ranges from less than 1.0 to 4.6 times greater than among never smokers (median approximately 1). The relative risks of smoking do not consistently differ in premenopausal and postmenopausal age groups. In addition, there is little consistency regarding the change in risk observed after smoking cessation. Former smokers have lower risks in some studies, but higher risks in others. Adjustment for other breast cancer risk factors does not appear to completely remove the weak association observed in some studies (Schechter, Miller, Howe 1985; Rohan and Baron 1989). In one study it was found that smokers tended to have a greater prevalence of tumor-positive axillary lymph nodes at the time of diagnosis than did never smokers and former smokers, a finding that could not be explained by patient delay (Daniell 1988). This association was not confirmed, however, ina recent report based on 10-year followup of the Nurses Health Study cohort that included 1,373 cases with information on extent of disease at diagnosis (London et al. 1989). This review of breast cancer and cigarette smoking suggests that cigarette smoking is not associated with breast cancer. Consistent changes in risk are not observed with smoking cessation. Endometrial Cancer The relationship between cigarette smoking and cancer of the endometrium is unique among the associations of smoking with cancers at various sites; of the sites for which smoking has been associated with a change in risk, endometrial cancer is the only cancer for which there is fairly consistent evidence of an inverse (protective) relationship (Baron 1984; Lesko et al. 1985; Stockwell and Lyman 1987), an effect that may be limited to postmenopausal women (Smith, Sowers, Burns 1984; Koumantaki et al. 1989). The reasons for the lower risk among women who smoke are not well under- 169 OL TABLE 7.—Studies of breast cancer and smoking cessation Risk relative to never smokers Yr Location (yr of Design Menopausal Current Former since Reference data collection) (number of subjects) status smokers smokers quitting Comments Cederlof et al. Sweden Prospective Pre 0.6 0.4 NR Cancer incidence (1975) (1963-72) (27,700) and post Schechter, Miller, Canada Case:contro] Howe (1985) (1980-82) (49:134) Pre 4.6 1.8 2} Adjusted for several breast cancer (71:2 19) Post Ll 0.8 2! risk factors Hiatt and Fireman Northern California Prospective Pre 1.2 1.2 NR Cancer incidence (1986) (1964-80) (84,172) Post lt 1.3 NR Brinton, Schairer, United States Case:control Stanford et al. (1973-75) (447:503) Pre 1 1.4 NR (1986) (614:818) Post Ll 1.0 NR Stockwell and Florida Case:control Pre 1.3° 0.9 NR Lyman (1987) (1981) (4,011:2,952) Post 1.2" 0.9 NR Brownson et al. Missouri Case:control (1988) (1979-86) (1142208) Pre 2.3 1.2 NR (206:872) Post 1.2 0.7 NR Adami et al. Sweden and Norway Case:control Pre 1.0 0.8 Relative risk calculated from (1988) (1984-85) (422:527) and post crude data Rohan and Baron Australia Case:control (1989) (1982-84) (146: 132) Pre 1.3 2.4 2] Adjusted for several breast cancer (280: 288) Post 1.5 09 2t tisk factors TL1 TABLE 7.—Continued Risk relative to never smokers Yr Location (yr of Design Menopausal Current Former since Reference data collection) (number of subjects) status smokers smokers quitting Comments London et al. United States Prospective Pre 1.0" 1d NR (1989) (1976-80) (117,557) Post 1 hl NR NOTE: NR=not reported. “Computed as a weighted average from cigarette dose-specific relative risks presented in the paper. Weights are the number of controls within each stratum of smoking. stood, but may be due to smoking effects on estrogen production and metabolism, including increased 2-hydroxylation of estradiol in smokers (Michnovicz et al. 1986), an earlier age at menopause in smokers (Baron 1984), and indirect effects of the body weight differences between smokers and nonsmokers, such as the production of estrogens from precursors within adipose tissue (MacDonald et al. 1978: Chapters 8 and 10). Table 8 includes a summary of findings from studies of endometrial cancer that have examined cigarette smoking cessation. Although the risk of endometrial cancer among current smokers in these studies ts approximately 30 percent lower than that among never smokers, the risk among ex-smokers is similar to, or slightly greater than. that among current smokers. This review of past research on endometrial cancer risk and cigarette smoking cessation suggests that current smokers are at lower risk of endometrial cancer than never smokers, but it is not clear whether this protective effect of smoking on endo- metrial cancer risk might be reversed soon after cessation of cigarette smoking. Although further investigation of the mechanisms for the protective effect of smoking on endometrial cancer is of scientific interest to better understand the effects of smoking on hormones and of hormones on endometrial cancer risk, this inverse association with smoking has no public health relevance, as the well-substantiated risks to other organ systems from continued smoking far outweigh any potential benefits to the endo- metrium. Other Cancer Sites The metabolic products of tobacco smoke can be found in ovarian follicular fluid (Hellberg and Nilsson 1988). However, there is little evidence that smoking is as- sociated with cancer of the ovary (Byers et al. 1983: Baron 1984; Baron et al. 1986: Stockwell and Lyman 1987; Whittemore et al. 1988: Mori et al. 1988). The risk of ovarian cancer differs little for either current or former smokers, as indicated in the only two studies that have examined the effect of cigarette smoking cessation on ovarian cancer risk (Table 8). Tobacco has been regarded as a contributing factor for cancer of the kidney (US DHHS 1982. 1989). The U.S. Veterans Study (Kahn 1966: Rogot and Murray 1980) and ACS CPS-IE (ACS, unpublished tabulations) suggest only small differences in mortality from renal cancer between current and former smokers (Table 8). A study of renal pelvis and ureteral cancers in Copenhagen (Jensen et al. 1988). however. showed a pattern of risk diminution with abstinence similar to that observed in bladder cancer. a site with the same histologic type of transitional-cell tumors. Cancers of the anus and penis are considered possibly to result from infection by a sexually transmitted agent in a way analogous to cancer of the uterine cervix (Daniell 1985; Daling et al. 1987: Hellberg et al. 1987). Smokers have been found to be at increased risk both for cancer of the penis (Hellberg et al. 1987) and anus (Daling etal. 1987: Holmes et al. 1988) in recent studies. Only one study has examined the effect of cessation on the risk of these cancers (Hellberg et al. 1987). This study found that 172 tdi TABLE 8.—Studies of cancer at selected sites that have examined the effect of smoking cessation Risk relative to never smokers Yr Population (yr of Design Cancer Current Former since Reference data collection) (number of subjects) site smokers smokers quitting Comments Cederlof et al. Sweden Prospective Endometrium 0.5 1.6 NP Cancer incidence (1975) (1963-72) (27,700) Lesko et al. 8 North American cities Case:contro! Endometrium 0.8" 09 21 Adjusted for obesity and (1985) (1976-83) (508:706) exogenous estrogens Stockwell and Florida Case:control Endometrium 0.8" 0.6 NP Lyman (1987) (1981) (990:2,952) Cederlof et al. Sweden Prospective Ovary 0.5 1.6 NP Cancer incidence (1975) (1963—72) (27,700) Stockwell and Florida Case:control Ovary ti 0.9 NP Lyman (1987) (1981) (696:2,952) Franks et al. United States Case:control Ovary | 0.9 >I Adjusted for age. parity, and use of (1987) (1980-82) oral contraceptives Kahn US veterans Prospective Kidney lt Ls NP Excludes “doctor's orders" quitters (1966) (1954-62) (248,195) Cancer mortality Rogot and Murray US veterans Prospective Kidney ld 1.2 NP Extension of US Veterans Study (1980) (1954-69) (293,958) Jensen et al. Copenhagen Case:control Renal pelvis 3.7 19 NP Crude relative risks computed (1988) (1979-82) (96:288) and ureter from dati presented PLt TABLE 8.—Continued Risk relative to never smokers Yr Population (yr of Design Cancer Current Former since Reference data collection) (number of subjects) site smokers smokers quitting Comments Hellberg et al. Sweden Case:control Penis 1.6 L.7 NP (1987) (NP) (244:232) Cederlof et al. Sweden Prospective Liver 2.4 1.0 NP Cancer incidence in males (1975) (1963-72) (27.300) Rogot and Murray US veterans Prospective Liver 2.3 1.8 NP Cancer mortality (1980) (1954-69) (248,000) Yuetal. Los Angeles. CA Case:control Liver 1s" a NP Abstainers for 210 yr were (1983) (1975--79) (76:76) considered never smokers Kahn United States Prospective Stomach 14 tl NP Excludes “doctor's orders” quitters (1966) (1954--62) (248,195) Cancer mortality Cederlof et al. Sweden Prospective Stomach 1.3 0.7 NP Cancer incidence in males (1975) (1963-72) (27,300) Rogot and Murray US veterans Prospective Stomach 1.5 1.1 NP Extension of US Veterans Study (1980) (1954-69) (293,958) Nomura et al. Japanese men in Hawaii Prospective Stomach 2.7 1.0 NP Cohort identified 1965-68 and (1990) (1965-68) (7,990) followed through October 1986 Kahn US veterans Prospective Leukemia 1.4 1.5 NP Excludes “doctor's orders” quitters (1966) (1954-62) (248,195) Cancer mortality SLI TABLE 8.—Continued Risk relative to never smokers Yr Population (yr of Design Cancer Current Former since Reference data collection) (number of subjects) site smokers smokers quitting Comments Cederlof et al. Sweden Prospective Leukemia (1975) (1963-72) (27,300) (Males) Ll 0.8 NP Cancer incidence (27,700) (Females) 0.4 1.0 NP Rogot and Murray —_ US veterans Prospective Leukemia 1.6 1.5 NP Extension of US Veterans Study (1980) (1954-69) (248,000) Trichopoulos et al. Greece Case:control Liver (1987) (1976-84) (104:454) HBxAg 3.3" 2.8 NP (89:454) HB,Ag* 1.6" 1.3 NP ACS CPS-II United States Prospective Kidney (unpublished (1982-86) (421,623) (Males) 2.7 2.1 NP Cancer mortality tabulations) (605,758) (Females) 1.5 1 NP NOTE: NP=not provided: HB,Ag=hepatitis B surface antigen; ACS CPS-H=American Cancer Society Cancer Prevention Study IL. “Computed as a weighted average from cigarette dose-specific relative risks presented in the paper. Weights are the number of controls within each stratum of smoking