Tobacco Documents Online

RESEARCH PAPERS Smoking habits and risk of cancers other than lung cancer: 28 years' follow-up of 26,000 Norwegian men and women Anders Engeland, Aage Andersen, Tor Haldorsen, and Steinar Tretli (Received 13 December 1995; accepted in revised form 26 February 1996) The impact of tobacco smoking on lung cancer risk has been investigated thoroughly since the 1950s, but other types of cancer also have been assodated with smoking. In the present study, the aim was to explore the variation in risk connected with cigarette, cigar, and pipe smoking of suspected smoking-assodated cancers other than lung cancer. Data were obtained from a survey of a random sample of the Norwegian population. A self-administered mailed questionnaire, which included questions about smoking habits, was completed by 26,000 men and women in 1965 (response rate: 76 percent). The cohort was followed from 1966 through 1993, including registration of all inddent cancer cases. A dose-response relationship of dgaret~e smoking to the risk of urinary bladder cancer and cancers of the upper digestive and respiratory tract was observed. For the latter forms of cancer, a dnse-response rehtionship of pipe smoking also was observed. In cancer of the pancreas, a stronger assodation between cigarette smoking and cancer risk was observed when the analysis was confined to histologically confirmed cases only. Current cigarette smokers at baseline had a significantly higher r/sk of cervical cancer than those who never smoked dgarettes. In cancers of the stomach, colon, rectum, breast, corpus uteri, ovar~ and prostate, and in leukemia, no association between smoking and cancer risk was observed. Cancer Causes and Conwol, 1996, 7, 497-506 Key words: Cancer incidence, cohort study, Norway, smoking. Introduction The impact of tobacco smoking on lung cancer risk has been investizated thoroughly since the 1950s. However, other forms of cancer also have been associated wlt~ smoking. The International Agency for Research on Cancer (IARC) evaluated studies on the carcinogenic risk of tobacco smoking in 1986.' In addition to lung cancer, tobacco smoking was regarded as an important cause of cancers of the urinary bladder, pancreas, and renal pelvis and also of oral, pharyngeal, esophageal, and laryngeal cancers. Although an increased risk of cancer of the uterine cerwx had oeen observed ~ ~mo~crs, ~ ,¢~o ~ concluded that the association was causal. It also was noted that tobacco smokers appeared to have an increased risk of cancer of the lip. In some studies, a reduction in the risk of endometrial cancer had been observed. How= ever, it was not concluded that smoking protects against Authors are zoith the Cancer Registry of Norway,. Odo, Norway. Address correspondence to Mr Engeland, The Cancer Reglsrry of Norway, Institute for Epidemiological Cancer Research, Montehello, N-0310 Oslo, Norway. This work was supported by grant no. 95080/001 from the Norwegian Cancer Society and Contract PH-64-499 from the US National Cancer Institute. 1996 Rapid Science Publishers C.a~er Cause* and Control Vol 7. t996 497 This a~acle/s for ind/vidual use oaly and may not be further repnxluced or st~red electronically ~0aout written pennissio~ from the copyright holder, Urmuthofiz~ reproduction may res~lt in F, nand at and o~her t:~-~lifi ca. (c) KLUWER ACADEMIC PUBL

| ! ! m A. Engeland e~ al cancer at this site. For cancers of both the stomach and liver, the available data did not permit a conclusion of causal relationship. More recently, it also has been claimed that leukemia should be added to the list of smoking- related diseases.2 The risk of frequent cancer types like cancers of the colon, rectum, breast, ovary, and prostate have been dis- cussed in relation to tobacco smoking in some studies.~ However; the evidence of an association between tobacco smoking and these forms of cancer is weak and inconsistent. In Norway, the tobacco smoking habit has been more widespread in men than in women. The difference between the genders in the proportion of smokers was large ha the 1960s, but has since diminished. In 1960, 64 percent of the Norwegian men and 27 percent of the women were smokers.7 The corresponding numbers in 1980 were 46 and 39 percent. In the present study, a cohort of 26,~0 Norweigan men and women who completed a questionnaire about smoking habits in 1965, was followed from 1966 through 1993. The impact of smoking habits on lung cancer risk recently has been explored in this datasct.* The aim of the present study was to explore the variation in risk con- nected with cigarette, cigar, and pipe smoking of other suspected smoklng-associated cancers. First, extensive analyses were performed for cancers accepted as tobacco- related ha the IARC monograph: cancers of the urinary bladder, kidney, pancreas, and a group of cancers in the upper digestive and respiratory tract. A similar analysis also was performed for cancer of the uterine cervix as this disease has been linked consistently to tobacco smoking. In addition, the risk in current and former smokers was compared with the risk in never-smokers for cancers of the stomach, colon, rectum, breast, corpus uteri, ovary~ and prostate, and for leukemia. No analysis was performed for cancers of the lip and liver due to the small number of cases. Material a.nd methods The data used ha the present analysis are from the Nor- we#an parts of the Migrant Study,* a collaborative study of diseases among migrants and native populations ha Great Britain, the United States, and Norway. A sample o~ 8,~0~ men born l~J-l~2/ an(a women born 1893-1932 was drawn randomly from lists of residents in Norway from the population census of 1960. The sampling fraction was 1/64 for persons born 1893- 1912 and 1/128 for persons born after 1912. An additional 2,554 men and 2,641 women were drawn from four selected counties (Ost-Agder, Vest-Agder, Rogaland, and Firmmark). Finally, a sample of siblings living in Norway of Norwegian-born rkaigrants to the US was included. The sibling sample included 6,127 men and 6,984 women. 498 Cancer Came~ ~nd Control. Vol 7. 1996 A self-administered questionnaire was mailed to all participants (37,380) during the Fall of 1964 to the Spring of 1965. For cigarette smoking habits, questions were asked about smoking status (never/former/current), age at the start of smoking, and the maximum number of cigarettes smoked for a minimum of one year. Former cigarette smokers were asked about the year of cessation. In addition; present cigarette smokers were asked about Wpe of cigarettes used (factory-made, with or without filter, and handrolled). The daily number of factory-made cigarettes and the weeldy number of packs of tobacco for handrolling (about 50 cigarettes per pack) were recorded. Questions about the use of cigars and pipe tobacco were restricted to smoking status (never/former/current), and the amount smoked in current smokers. In addition to smoking habits, the questionnaire also inquired about cardiovascular and respiratory symptoms.'° Of the selected participants, a total of 1,485 persons died or emigrated before 1 January 1966. Seventy-slx percent of the remaining persons completed the question- naire (27,136). Thereafter, 623 persons were excluded due to identification problems. Finally, 381 who did not answer any of the questions about cigarette smoking habits were excluded. A total of 26,132 persons were left for analysis, 11,863 men and 14,269 women. Since 1964, all residents of Norway have had a unique D-digit individual identification number. Not only does this number make the identification of individuals simple and reliable, it .also makes it possible to link registry data with other data sources. All deaths and emigrations of Norwegian citizens are registered continually in the Central Population Register of Statistics Norway, based on the individual identification number. Since 1953, the Norwegian Cancer Registry has received information on all cancer patients in the entire population. The reporting system is based on pathology" and cytology reports, clinical reports, and death certifi- cates. The reporting of cancer cases is compulsory for physicians. Site, histologic type, stage of disease at the time of diagnosis, residence, and the 11-diglt individual identification number are reported. The coding system is based on a modified version of ICD-7?' Separate analyses were performed for cancers of the urinary bladd~ (ICD-7 code 181), kidney (code 180), pancreas tcoae to/), and me uterine cervix tcode lit), in addition to a group of cancers of the upper digestive and respiratory tract (oral cavity, pharynx, esophagus, and larynx; ICD-7 codes 141, 143-148, 150, and 161, respec- tivdy). In each analysis, persons with a diagnosis of the cancer in question prior to 1 January I966 were excluded. All persons included ha the analysis were followed from 1 January 1966 to the date of the first diagnosis of the type of cancer in question, the date of emigration, the date of

Smoking habits and cancer risk deat~ or to the end of 1993. The only exception was in the a~n~lysis of cancers o~ the upper d~gesdve ~d re~r- ato~ ~c~ ~ t~s ~oup o~sites, more ~ one d~nosis per p~on wm ~lowed. Howler, o~y ~ I~st ~osis at ~ mato~c site ~s ~duded. At ~e st~ o~ ~ollow- up, ~e men were aged 3~ to 72 y~m md ~e women were aged 33 to 72 y~m. ~ both g~dem, ~e me~m age ~ 56 yems. ~e mated~ ~s ~vided by cig~ett~smo~g habirs at bm~e hto n~er, foyer, md cu~ent smokers. nev~-smokers conslmed of ~ose who had never smoked ci~e~, plus ~ose who did not smoke at the ~te~ew but ~ ~o~ pre~ous ci~re~e smo~g. Cu~ent dg~ette smokem at bmeline were ~tegofiz~ accor~g to d~ly conmmpdon ~to four 6oups: 14, .5-9, 1~14, md 15+ ci~e~ per day. In the f~st ~oup, oc~on~ smokers smo~g l~s ~an one cigare~e a day were ~duded. In ~e ~ys~, some of ~e latter groups. were comb~ed due to sma~ numbers. Some smokers stop smo~g when ~ey e~efience symptoms of a serious tobacco-related ~sease. For ex~pl~ foyer smokers may e~efience a ~gher risk of l~g ~cer ~m cument smokes in the c~don.TM In ~I malyses, persons were reg~ded as cu~ent dg~e~e smokes unt~ five ye~s alter cessation of cig~e~e smo~g, md ther~er as foyer smokem. Cu~ent ci~e~e smokes at b~e~ne were chss~ accord~g to age at the st~ of smo~ng ~to t~ee categoric: yo~ger thin 20 yems, 20 to 29 y~m, md 30 yems or older. ~ey flso were ~tegodzed by ci~ette smoked hto four groups-~e., factor-made, hm~olle~ bo~ hm~o~ed md h~o~-made cigmett~, ~d ~o~. For bo~ ci~ md pipe smo~g, go~don wm used: never, ~o~er, or current smokem of cigar or pipe, respectively. Never-smokem ~clud~ those with unkno~ habits o~ cigar/pipe smo~ng. Cu~ent smokers at base~e were ~tegofized accor~g to ~ly conmmp~on (cig~ 1~ g ~d 5+ ~ pip~: 1~ g, 5-9 g, md 10+ g). ~ pemons ~ the coho~ were ~tegofiz~ by place of r~idence ~ 1960 as H~g in urbm or ~l muddp~des accor~g to a de,don ~ven by S~dstics Nosy. EPIC~.~ Relative ds~ ~) md their 95 perc~t co~dence ~tew~s (C~ were derived from mnltivmht~ Cox pro~on~ ~ds r~don m~s ~ a~ age m ~e ~e variable J* Models were fi~ed ~or men ~d women s~tely. Exert ~or a v~able chma~efiz~g the bmel~e cigmette co~pdo~ ~e o~v~l~ 0ge at ~e m~ oI smo~g ci~e~es, ~pe o~ cigare~e, pipe smo~g, cig~ smo~g, ~d urb~ place o~ r~idence) were ~cluded model only ~ ~ey proved si~Ii~t ~ven consumption. When other main effects than cigarette smoking were included in the model, first-order interac- tion terms were tested. A less extensive analysis was performed for cancers of the stomach (ICD-7 code 151), colon (code 153), rectum (code 154), breast (code 170), corpus uteri (code 172), ovary (code 175), and prostate (code 177), and for leuke- mla (code 204). Only one smoking variable with three categories was included in this model: never, former, and current smoker. Throughout the analysis, a significance level of five percent was used. Results During the follow-up, the total observed person-time was about 230,000 person years (PY) in men and 310,000 PYs in women. A detailed list of the PYs included in the present study are tabulated in another paper.8 The PYs were slightly different for the separate forms of cancer in the present study due to differences in follow-up; subjects were not followed after their first diagnosis of the cancer in question. At baseline, 17 percent of the men were never-smokers, 66 percent were current smokers, and a further six percent had stopped smoking less than five years prior to baseline. In women, 68 percent had never smoked, six percent were former smokers, 24 percent were current smokers, and a further one percent had stopped smoking less than five years prior to baseline. Cancer of the urinary bladder A total of 307 cases of cancer of the urinary bladder (95 percent histologically verified) were registered during the follow-up, 221 in men and 86 in women. Current cigarette smokers at baseline combined had an RR of 32. (CI = 2.1-4.7) compared with never-smokers of cigarettes in men, and a corresponding RR of 2.8 (CI = 1.8-4.3) in women. A clear dose-response relationship in the risk of urinary bladder cancer by cigarette smoking was revealed in both genders (Table 1). In men, the RR in current cigarette smokers at basellne of one to four cigarettes a day compared with never-smokers of ciga- rettes was 2.5, while the RR in those smoking 15 cigarettes or more a day was 5.1. In women, the RR rose from 1.5 in those smoking one to four cigarettes a day to 7.9 in those smoking 15 cigarettes or more a day. In both genders, the former smokers had RRs at about the same level as current smokers at baseline with a daily consump- tlon of one to four cigarettes. No significant effect was seen for age at the start of smoking, type of cigarette, cigar or pipe smoking or place of residence. Cancer of the kidney A total of 147 cases of kidney cancer (85 percent C~cer Cau_,e~ a~d ControL Vol 7. I9% 499

A. Engeland et al Table 1. Number of cases and relative risks (RR) of urinary bladder cancer with 95 percent confidence intervals (CI) Men Women No. RR (~1) No. RR (CI) Cigarette smoking Never 32 1.0 Referent 47 1.0 Referent Former 62 2.1 (1.3-3.2) 6 1.5 (0.6-3.5) Current (at baseline) 1-4 clg/day 31 2.5 (1.5-4.0) 8 1.5 (0.7-3.2) 5-9 cig/day 27 2.7 (1.6-4.5) 8 2.2 (1.0-4,7) 10-14 clg/day 38 3.4 (2.1-5.4) 11 5.4 (2.8-11) >- 15 cig/day 30 5.1 (3.1-8.4) 6 7.9 (3.3-19) Unknown consumption 1 9.1 (1 2-67) 0 -- -- Table 2. Number of cases and relative risks (RR) of kidney cancer wi~ 95 percent confidence intervals (CI) Men Women No. RR (CI) No. RR (CI) Cigarette smoking Never 19 1.0 Referent Former 28 1.3 (0.8-2.4) Current (at baseline)a 0-4 clg/day 8 0.9 (0.4-2.1) 5-9 cig/day 13 1.8 (0.9-3.6) >- 10 cig/day 19 1.3 (0.7-2.5) Unknown consumption 0 -- -- Residence Rural 47 1,0 Referent Urban 40 1,8 (1.2-2.8) 45 1.0 Referent 1 0.2 (0.0-1.7) 14 1.1 (0.6-2.0) Numbers for women were too small to group according to amount of consumption, histologically verified) were registered during the follow- up, 87 in men and 60 in women. Current cigarette smokers at baseline had art RR of 1.4 (CI = 0.8-2.5) compared with never-smokers of cigarettes in men, and a corresponding RR of 1.1 (CI = 0.6-2.0) in women. In men, a hlghe~; although not significant, risk was observed in current smokers at baseline of five cigarettes or more a day compared with never-smokers of cigarettes. (Table 2). A signiflcandy higher risk was found in men living in urban areas compared with men living in rural areas. Due to the small number of cases, the female current smokers at baseline were not grouped according to consumption. No significant effect was seen cigar or pipe smoking. Cancer of the pancreas A total of 224 cases of cancer of the pancreas were reg- istered during the follow-up, 109 in men and 115 in women. Only 55 percent of the cases were histologically verified. Due to the large number of cases not histologically confirmed, analyses including histologically confirmed cases only were performed in addition to the analyses including all cases. Current cigarette smokers at baseline had an RR of 1.2 (CI = 0.8-1.9) compared with never-smokers of clgarettes in men, and a corresponding RR of 1.4 (CI = 0.9-2.1) in women. The risk in current smokers at baseline increased with consumption (Table 3). No significant effect was seen for age at the start of smoking, type of cigarette, cigar, or pipe smoking, or place of residence. In an analysis restricted to histologically confirmed cases, a sharper increase in the risk by consumption in current dgarette smokers at baseline was observed in men (Table 3). In women, only small differences were seen. Cancer of the upper digestive and respiratory tract A total of 126 cases of cancer of the upper digestive and respiratory tract (94 percent histologically verified) were registered during the follow-up, 87 in men and 39 in women. In men, both cigarette and pipe smoking were included in the model. In addition, an interaction term between cigarette and pipe smoking was included. Current smok- ers of both cigarettes and pipe at baseline who smoked 500 Cancer Came= and Control. Vol 7. 1996

Smoking habits and cancer risk Table 3. Number of cases and rolatlvo risks (RR) of cancar of the pancreas with 96 percent confldsnca Intervals (Cl) All cases Men Histologically vedfled cases only Women All cases No. RR (CI) No. RR (Cl) No. RR (CI) Cigarette smoking Never 31 1.0 Referent 12 1.0 Referent 82 1.0 Referent Former 28 0.9 (0.6-1.5) 16 1,3 (0.6-2.8) 4 0.6 (0.2-1.5) Current (at baseline) 1-4 cig/day 12 0.9 (0.5-1.8) 5 0.9 (0,3-2.7) 8 0.9 (0.4-1.8) 5-9 cig/day 11 1.0 (0.5-2.1) 6 1.4 (0.5-3.7) 0 N _ >_ 5 cig/daya ...... 21 1.8 (1.1-3.0) 10-14 cig/dayb 16 1.3 (0.7-2.4) 11 2.1 (0.9-4.9) -- -- -- ~ 15 cig/dayb 10 1.6 (0.8-3.2) 8 2.9 (1.2-7.1) -- -- -- Unknown consump~on 1 7.9 (1.1-58) 0 -- -- 0 -- N a Women only. b Men only. Table 4. Number of cases and relative risks (RR) of the upper digestive and respiratory tract with 95 percent confidence intervals (Ci)a Men Women No. RR (Cl) No. RR (el) Cigarette smoking Never 19 1.0 Referent Former 15 0.5 (0.3-1.1) Current (at baseline)b 1-4 cicj/day 12 1.2 (0.6-2.7) 5-9 cig/day 9 1.1 (0.6-2.7) 10-14 c~g/day 16 1,8 (0.9-3.8) >_ 15 cig]day 16 5.4 (2.5-12) Unknown consumption 0 -- -- Pipe smoking Never 18 1,0 Referent Former 22 1,3 (0.7-2.6) Current (at baseline) 1-4 g/day 18 3.1 (1.6-6.2) 5-9,g/day 12 3.1 (1.4-6.6) >_. 10 g/day 15 8.7 (4.0-19) Unknown consumption 2 7.5 (1.7-33) 26 1.0 Referent 1 0.5 (0.1-3.4) 12 1.9 (0.9-3.8) Interaction term between cigarette and pipe smoking not shown. Numbers for women were too small to group according to amount of consumption. more than 14 cigarettes a day or more than nine grams ha a pipe, had a significandy lower RR than the product of the RRs for the group of those smoking more than 14 cigarettes, and the group of those smoking more than nine grams ha a pipe. The aim of this analysis was to estimate the main effects of each factor involved adjusted for the other factors. Consequently, the interaction terms ~e not shown. In men, a dose-response relafionsl~p w~s observed for both cigarette and pipe smoking (Table 4). No significant effect of age at the start of smoking, type of cigarette, cigar, or place of residence was seen. In women, only three categories of cigarette smoking were included in the model due to small number of cases: never, former, and current smokers at baseline. Cancer of the uterine cervix A total of 86 cases of uterine cervical cancer (99 percent histologically verified) were registered during the follow- up. Current cigarette smokers at baseline had an RR of cervical cancer of 2.5 (CI = 1.6-3.9) compared with never- smokers of cigarettes. In Table 5, the current smokers at baseline were divided ires categories according to dally cigarette consumption. Those smoking less than five ciga- C~ncerCau~e~and ControL VolT. 1996 501 0

A. Engeland et al Table 5. Number of cases and relative risks (RR) of cancer of the uterine cervix with 95 percent confidence intervals (C[) Wonlen No, RR (CI) Cigarette smoking Never 39 1.0 Referent Former 5 1.0 (0.4-2.6) Current (at baseline) 1-4 clg/day 12 1.9 (1.0-3.6) 5-9 cig/day 20 3.3 (1.9-5.8) > 10 cig/day 10 2.4 (1.2-4.8) Unknown consumption 0 w -- rettes had the lowest risk, but those smoking five to nine cigarettes appeared to have a higher risk than those smok- hag 10 or more cigarettes a day. No difference ha the risk was observed between former and never-smokers of ciga- rettes. Age at the start of smoking, type of cigarette, cigar, or pipe smoking or phce of residence had no significant effect on the risk of cervical cancer. OtJ~er cance~s For c~ncers o~ the stomach, colon, recuLm, breast, corpus uteri, ovary; ~d prostate, ~d for leuke~ the n~b~ of ~ ~d ~s o~ breeze cu=ent ~d foyer smokem comp~ed ~th n~er-smokem me =bulated h Table 6. ~e propo~on of ~ses ~stolo#~y vexed r~g~ ~om 84 to 99 percent h ~e e~cers. ~e ~sk for ~y of ~ese disemes h current ~d foyer smokem was not sightly ~erent from ~at in n~er-smokers. ~e ~k of re~ ~cer h ~t smokers m bmelhe had ~ ~ of 1.6 of border~e si~ic~ce ~ men, whereto, ~ wome~ ~e ~ wm 0.8. For comp~so~ s~ ~- ~t~ for c~cers of ~e u~ bladde~ ~e~ pancr~, upper ~g~dve =d resp~ato~ tmcg ~d ute~e ce~ ~e ~d~ ~ Table 6. Discussion" In this study, the risk of cancers at different sites connected with tobacco smoking was explored in a cohort of 26,000 Norwegian men and women followed for 28 years. The strongest dose-response relationship for cigarette smoking was found for urinary bladder cancer and cancers of the upper digestive and respiratory tract. Also, a dose- response relationship for pipe smoking was found for the latter forms of cancer. The study cohort was mainly a stratified sample with unequal sampling fractions from the Norwegian popula- tion. The main objective of the analysis was cancer incidence at given smoking habits. It is difficult to know whether there is a different association between smoking 502 Cancer Cause= and Control Vol 7. 1996 and cancer incidence among the nonrespondents. The personal identification was known for 53 percent of these. As a group, they had a higher risk of lung cancer than the respondents after adjustment for place of resldence3 .This effect may be due to a higher proportion of smokers among the nonrespondents. Smoking habits were recorded only once in this study -- in 1964-65. Self-reported smoking status (smoker or nonsmoker) is regarded as quite accurate,ix*6 Since the persons were between 31 and 72 years of age at interview, relatively few of the never-smokers are likely to have started smoking afterwardsYlt is more likely that current and former smokers at the time of interview would change their smoking habits in subsequent years. Former smokers might have started to smoke again, resulting in an over- estimation of the RR in former smokers. However," comparisons of the RRs in former smokers and those in baseline current smokers suggest that relatively few former smokers have become smokers again after being interviewed. In addition, baseline current smokers might have stopped smoking or changed their daily consump- tion. A reduction in cigarette consumption would resuk in a bias towards unity in the estimation of the RR con- nected with smoking. A similar bias would be introduced if baseline current smokers stopped smoking during the 28 years of follow-up. According to Renneberg et al,*z the proportion of daily smokers in the cohorts included in the present study decreased a~ter 1965 in Norway. For example, among those born 1910-14, 61 percent of the men and 22 percent of the women smoked in the age bracket of 50 to 54 (around 1965). At ages 70 to 74, 34 percent of the men and 10 percent of the women were smokers. Self-reporting of the number of cigarettes smoked per day is problematic.16 Smokers tend to self-report a multiple of five or 10 cigarettes. In the present study, the daily number of factory-made cigarettes and the weekly number of packs of tobacco for handrolllng (about 50 cigarettes per pack) were registered. Peaks were seen at multiples of five cigarettes for factory-made cigarettes and at half-packs for handrolled cigarettes. Information on the number smoked was used in categorical variables only, but some mlsclasslficatlons may have occurred. Among those termed 'never-smokers' (of cigarettes), 32 percent of the men and 29 percent of the women did not answer the question on former cigarette smoking. However; no difference in lung cancer risk was seen in these persons compared with those known to be never- smokers at the time of interview? Among those termed 'never-smokers' of cigars, 32 percent of the men and 35 percent of the women did not answer any question on cigar smoking. Among those termed "never-smokers' of pipes, 30 percent of the men and 42 percent of the women did not answer any question on pipe smoking. None of

Smoking habits and cancer risk Table 6. Number of cases and relative dsks (RR) of different types of cancer wit~ 95 percent confidence intervals (CI) in baseline former and current smokers Site Men Women No. RR (Cl) No. RR (CI) Upper digestive and respiratory tract Never smoker 6 Former smoker 5 Current smoker 76 Stomach Never smoker 39 Former smoker 50 Current smoker 169 Colon Never smoker 41 Former smoker 39 Current smoker 150 Rectum Never smoker 20 Former smoker 16 Current smoker 103 Pancreas Never smoker 17 Former smoker 19 Current smoker 73 Breast Never smoker Former smoker Current smoker Cervix uted Never smoker Former smoker Current smoker Corpus uted Never smoker Former smoker Current smoker Ovary Never smoker Former smoker Current smoker Prostate Never smoker 139 Former smoker 117 Current smoker 451 Kidney Never smoker 12 Current smoker 57 Urinary bladder Never smoker 21 Former smoker 34 Current smoker 166 Leukemia Never smoker 16 Former smoker 14 Current smoker 34 1.0 Referent 26 1.0 Referent 0.9 (0.3-2.8) 1 0.5 (0.1-3.3) 3,5 (1,5-8.1) 12 1.8 (0.9-3.7) 1.0 Referent 119 1.0 Referent 1.3 (0.9-2.0) 9 0.8 (0.4-1.6) 1.3 (0.9-1.9) 31 1.0 (0.6-1.4) 1.0 Referent 211 1.0 Referent 1.0 (0.6-1.5) 26 1.3 (0.9-2.0) 1.2 (0.8-1.6) 63 1.1 (0.8-1.4) 1,0 Referent 104 1.0 Referent 0.8 "(0.4-1.6) 13 1.3 (0.8-2.4) 1,6 (1.0-2.6) 24 0.8 (0.5-1.3) 1.0 Referent 81 1.0 Referent 1.2 (0.6-2.2) 5 0,7 (0.3ol .7) 1.3 (0.8-2.2) 29 1.4 (0.9-2.1) 418 1.0 Referent 47 1.1 (0.8-1.5) 138 1.0 (0.8ol.2) 39 1.0 Referent 5 1.0 (0.4-2.6) 42 2.5 (1.6-3.9) 91 1.0 Referent 12 1.2 (0.6-2.2) 37 1.1 (0.7-1.6) 98 1.0 Referent 6 0.6 (0.2-1.3) 36 1.0 (0.7-1.5) 1.0 Referent 0.9 (0.7-1.1) 1.1 (0.9-1.3) 1.0 Referent 45 1.0 Referent 1.4 (0.7-2.6) 14 1.1 (0.6-1.9) 1.0 Referent 45 1.0 Referent 1.7 (1.0-2.9) 7 1.8 (0.8-3.9) 2.5 (1.6-3.9) 34 2.9 (1.9-4.6) 1.0 Referent 37 1.0 Referent 0.9 (0.4-1.9) 1 0.3 (0.0-2.2) 0.6 (0.4-1.2) 13 1.3 (0.7-2.5) Cancer Cause= and ControL Vol 7. 1996

these ~roups showed a higher risk of lung cancer than never-smokers. It therefore was assumed that mos~ of them had never smoked cigars or pipes, respectively. Cigarette smoking, pipe smoking, and cigar smoking were all coded as separate variables. However, when pipe or cigar smoking was not included in the model, models with never-smokers of any tobacco as the reference group were fitted (not shown). No differences in the ILK estl- mates from those presented for cigarette smoking were observed. The effect of tobacco smoking on lung cancer was explored in a recent article from the present dataset3 The dose-response relationship found for cigarette smoking, and also for pipe smoking in men, was in agreement with previously published studies. These results indicate that the information on smoking habits is quite reliable, and it should be possible to obtain information on effects on other smoklng-related cancers as well. However, the only information available on risk factors is smoking habits and place of residence (urban/rural). In other types of cancer where the proportion of the cases attributable to smoking is lower than for lung cancer, the influence of confounders may distort the results to a larger extent. Consequently, the presented size of the estimates should be interpreted with caution. Also, the number of cases included in the present study was limited for some types of cancer and the smoking categorization was crude. Con- sequently, small differences in risk in smokers compared with nonsmokers could be difficult to identi~. Land and Zelner-HenriksenTM have analyzed the impact of smoking on cancer risk using most of the cohort (persons born 1895 -1929) described above with follow-up from 1966 to 1977. In the present study, a much larger number of cases was included, allowing a more detailed analysis. Most studies have shown a two- to fourfold increase in risk of urinary bladder cancer1'19 in current smokers . compared with never-smokers. In two more recent popu- latlon-based ease-control studies,1°~1 a twofold and a fivefold increased risk was found in current cigarette smokers compared with never-smokers, respectively. The risk of urinary bladder cancer was not influenced by pipe or cigar smoking.~ A significant dose-response relation- ship of number of cigarettes per day to the risk of urinary study, similar results were found. Pipe or cigar smoking showed no effect on the risk of urinary bladder cancer, while a dose-response relationship was revealed for ciga- rette smoklno~ in, both genders. \Y/e .¢ound a t~rcefo!~ increased risk in current cigarette smokers at baseline compared with never-smokers of cigarettes. Cigarette smoking is the major risk factor identified for cancer of the kidney,= However, tobacco smoking has been regarded as an important cause only for cancers of so~ Cancer Came* and Control. Vol 7. 1996 the renal pelvis) Several studies have revealed an increased risk in cigarette smokers for renal cell carcinomas, although the increase has been srnaller than for cancers of the renal pelvis.~ In the present study, most of the cases were renal cell carcinomas (86 percent), and the association between tobacco smoking and risk of kidney cancer was weak. The increased risk found in men living in urban areas compared with men living in rural arms may be due to lifestyle factors other than smoking Tobacco smoking is the factor most consistently emerg- ing as a risk factor for pancreatic cancer." Studies have shown a two to three times higher risk among smokers than never-smokers.19;2 Heuch et a124 analyzed parts of the present cohort followed to the end Of 1978, combined with data from another study, where information on alcohol and coffee consumption was available. Informa- tlon on cigarette smoking was available for 38 cases in men. In the present study, the effect of tobacco smoking was small when all cases were included, and the RR of current cigarette smokers at baseline compared with never° smokers was not significant. However, as in Heuch et al, we found a clearer dose-response relationship when the analysis was confined to histologically confirmed cases. In the evaluation of the importance of smoking from the LARC (1986), it was not possible to conclude that the association between tobacco smoking and risk of uterine cervical cancer was causal.~ Nevertheless, an association between both prelnvaslve and invasive cervical cancer has been reported.2~ Two recent meta-analysis reviews2saz of studies of smoking and cervical cancer reported a sum- mary estimate of RR for the smoking effect to be 1.5 and 1.7, respectlvely. A dose-response relationship has been noted, and the highest risk generally has been observed for long-term smokers.~ Other studies, however, have indicared that the observed effect of smoking on cervical cancer may be due to confounding, even after adjustment for number of sexual partners.~ In the present study, a higher risk was found in current cigarette smokers at baseline compared with never-smokers, but the risk did not increase monotonically by cigarette consumption. Since no adjustment for sexual behavior and/or human papfllomavirus (HPV)-status was included in the analysis, the observed effect of cigarette smoking may be due to confounding. urinary bladder, and pancreas has been observed in ciga- rette smokers than in pipe or cigar smokers, the difference in the risk of cancers of the upper digestive and respiratory ma~i~cu, m the present srucxy, a dose- response relationship was revealed in both cigarette and pipe smoking in men. In a 1991 review of risk factors for stomach cancer,~ it was stated that evidence indicates that stomach cancer is associated with tobacco consumption. It also was noted

thah even in long-term heavy smokers, the risk did not exceed twice the risk in never smokers. The present study showed only a nonsignificant RR in men of 1.3 in baseline smokers compared with never-smokers. In women, no difference in risk was observed. In an Australian population-based case-control study~ on the risk of colorectal adenocarclnoma connected to tobacco smoking, an RR of 1.7 was found in males smoking handrolled c~garettes compared with nonsmokers. How- ever, no excess risk was found in all current smokers combined compared with nonsmokers. After a review of the literature, the authors concluded that there is insuf- ficient evidence to demonstrate an association between tobacco smoking and colorectal cancer. In the present study, no significant associations were found between tobacco smoking and the risk of colorectal cancers. The RE. of 1.6 found for rectum cancer in current smokers at baseline compared with never-smokers in men was near statistical significance at the chosen level. However, a similar relation was not seen in women. An association between smoking and early menopause has been shownJz Since the risks of cancers of the breast and corpus uteri increase with higher age at menopause, a lower risk for these cancer types may be expected in smokers than in nonsmokers. However, the expected reduction is, at least in breast cancer, quite small.* Both protective and harmful effects of smoking on breast cancer risk have been suggestedJ~2 Palmer and Rosenberg* concluded in a review that there was little evidence to suggest that cigarette smoking influences the risk of breast cancer. In the present study, no difference in risk was seen in baseline current, former, and never- smokers. A reduction in the risk of endometrial cancer in smokers has been found in several studies, although in some studies the opposite relation has been observed.~ In the present study~ no effect of smoking was observed. The itisk of ovarian cancer also increases with b.lgher age at menopause, and a lower risk in current smokers has been indlcated.5O~In the present study, no effect of smoking was observed on the risk of ovarian cancer. The epiderniologlc evidence of an association of smoking and risk of prostate cancer was reviewed by Nomura and Kolonel in 1991.6 Most studies have shown no association. ann mttte studies where an association was found, no dose-response reladonskip was observed. The authors concluded that cigarette smoking does not increase the risk of prostate cancer. In the present study, no association between smoking and prostate cancer was found. Due to the recent interest shown ha a possible associa- tion between tobacco smoking and leukemia, this cancer was included in the present study.~ A meta-analysis by Brownson et al ~7 supported a causal relationship between cigarette smoking and certain forms of adult leukemia. Sraoking habits ann cancer risk After a review of 15 epidemiologic studies (all included in the recta-analysis by Brownson et a/), Siegel~ concluded that leukemia should be regarded as a smoking-rehted disease. In the present study, no association between smoking and leukemia was found. However, the number of cases included was limited and the smoking categori- zation was crude; thus, a possible small excess risk in smokers would be di~cult to reveal. In conclusion, the present study revealed a dose- response relationship of cigarette smoking to the risk of urinary bladder cancer and cancers of the upper digestive and respiratory tract. For the latter forrns of cancer, a dose-response relationship for pipe smoking also was observed. In cancer of the pancreas, a stronger association between cigarette smoking and cancer risk was observed when the analysis was confined to histologically con- firmed cases only. In cancer of the kidney, only indications of an association between cigarette smoking and cancer risk were observed. Current cigarette smokers at baseline had a significandy higher risk of cervical cancer than never-smokers of cigarettes. In cancers of the stomach, colon, rectum, breast, corpus uteri, ovary, and prostate, and in leukemia, no association between smoking and cancer risk was observed. l. International Agency for Research on Cancer. Tobacco Smoking. Lyon, France: IARC, 1986; IARC Monogr £~al Cardnog Risk Cbem Hum, Vol. 38. 2. Siegel M. Smoking and leukemia: evaluation of a causal hypothesis. AmJ El~ideraio11993; 138: 1-9. 3. Kune GA, Kune S, "qitetta L, Watson LE Smoking and colorectal cancer risk: data from the Melbourne Colorectal Cancer Study and brief review of literature. Int J Cancer 1992; 5ft. 369-72. 4. Palmer JR, Rosenberg L. Cigarette smoking and the risk of breast cancer. E~ideraiol Re¢~ 1993; 15: 145-56. 5. Whlttemorc AS, Wu ML, P~ffenbarger RS jr, et at Personal and environmental characteristics related to epithelial ovar- ian cancer. II. Exposures to talcum powder, tobacco, alcohol, and coffee. Am.[ E~ideraio11988; 128: 1228-40. 6. Nomur~ AMY, Kolonel LN. Prostate cancer: a current per- spectlve. El~iderniolRe~ 1991; 13: 200-27. 7. M~rck HI, Linde J, Agner E, Hein HO, GyTatelberg F, Nielsen PE. Tobaksforbrug og rygevaner i Norden 1920- Nordic countries 1920-1~80] Nordlsk Medicin 1982; 97:134- 46. (In Danish) 8. Engeland A, Haldorsen T, Andersen A, Tredi S. The impact of smoking habits ort lung cancer risk: 28 y~ars' observation of 26,000 Norwegian men and women. Cancer Causes Control 1996; 7: 366-76. 9. Reid DD. Studies of disease among migrants and native populations in Great Britain, Norway and the United States. I. Background and design. Im Haenszel W) ed. Epidemiologi- cal Study o f Cancer and Other Cbronic Diseases. Washington DC: US Dept of Health Education and Welfare, Public Health Service, 1966; 287-90. Cancer Came~ and Control. Vol 7. 19%

A. Engeland et al 10. Haenszel W, Hougen A. Proportion of respiratory symp- toms in Norway.J Chron Dis 1972; 25: 519-44. 11. World Health Organization. International Classification of Diseases, Seventh Revision. Geneva, Switzerland: WHO, 1957. 1Z Halpem MT, Gillespie BW, Warner KE. Patterns of absolure risk of lung cancer mortality in former smokers.JNC11993; 8~: 457-64. 13. Preston DL, LublnJH, Pierce DA. EPICURE. User's Guide. Seattle, WA (USA): Hirosok International Corporation, 1988-93. 14. Cox DR, Oakes D.Analysis of Survi~alData. London, UK: Chapmarm and Hall Ltd, 1984. 15. Patrick DL, Cheadle A, Thompson DC, Diehr P, Koepscll T, Kinne S. The validity of sdLreportcd smoking: a review and recta-analysis. Am ] Public Heal~b 1994; 84: 1086-93. 16. Klesges RC, Debon M, Ray J~. Are self-reports of smoking rate biased? Evidence from the second national health and nutrition examination survey. J Clin Epidemiol 1995; 48: 1225-33. 17. Rormeberg A, Lurid KE, Hafstad A. Lifetime smoking habits among Norwegian men and women born between 1890 and 1974. Intf Epidemiol 1994; 23: 267-76. 18. Land E, Zeiner-Henriksen T. Reking sore risikofaktor for ulike kreffformer blant 26 000 norske mcrm og kvlnner. [Smoking as risk factor for cancer among 26,000 Norwegian males and females.] Tu:Isskr Nor Laegeforen 1981; 101:1937- 40. (In Norwegian) I9. US Department of Health and Human Services. Reducqng the health consequences of smoking: 25 years of progress. A report of the Surgeon General. Washington DC: DHHS, 1989. 20. Butch JD, Rohan TE, Howe GR, et aL Risk of bladder cancer by source and type of tobacco cxposur~ a case-con- trol study. IntJ Cancer 1989, 44: 622-8. 21. Momas I, Danres JP, Pesty B, Bontoux J, Gremy F. Bladder cancer and black tobacco cigarette smoking. Some results from a French case-control stud~ EurJ Epidemlo11994; 10: 599-604. 22. Tomatis L, Aitio A, Day NE, et aL Cancen Causes, Occur- rence and Control Lyon, France: International Agency for Research on Cancer, 1990; IARC Sol. Pub. No. 23. Mellemgaard A, Engholm G, McLanghlin JK, Olsen JH. Risk factors for renal cell carcinoma in Denmark. I. Role of socioeconomic status, tobacco use, beverages, and family history. Cancer Causes Control 1994; 5: 105=13. 24. Hcuch I, Kv~le G, Jacobsen BK, Bjelke E. Use of alcohol, tobacco and coffee, and risk of pancreatic cancer. BrJ Cancer 1983; 48: 637-43. 25. Gram IT, Austin H, Stalsberg H. Cigarette smoking and the incidence of cervical intraepithdial neoplasia, grade III, and cancer of the cervix uteri. Am J Epidemio11992; 135: 341-6. 26. Sood AK. Cigarette smoking and cervical cancer:, mera- analysis and critical review ofrecentstudies.AmJPrev Med 1991; 7: 208-13. 27. Licciardone JC, Brownson RCo Chang JC, Wilkins III JR. Utezine cervical cancer risk in cigarette smokers: a recta- analytic study. Am J Prey Med 1990; 6: 27,1-81. 28. Winkelstein W Jr. Smoking and cervical cancer-current status: a review. Am J Epidemio1199~ 131: 945-57. 29. Phillips AN, Smith GD. Cigarette smoking as a potential cause of cervical cancer: has confounding been controlled? IntJ Epidemio11994; 23: 42-9. 30. Doll R, Peto R. The causes of cancer : quantitative estimates of avoidable risks of cancer in the United States today.JNCI 1981; 66: 1191-308. 31. Forman D. The Etiology of Gastric Cancer. Lyon, France: International Agency for Research on Cancer, 1991; I.ARC Scl. Pub. No. 105: 22-32. 32. Baron JA. Smoking and estrogen-rdated disease. Am J Epidemiol 1984; 119: 9-22. 33. Brhaton LA, Barrett RJ, Bemaan ML, Mortel R, Twlggs LB, W'flbanks GD. Cigarette smoking and the risk of endo- inertial cancer. AmJ Epidemio11993; 137: 281-91. 34. Shu XO, Brinton LA, Zheng W, Gao ~ Fan J, Fraumenl ]'F Jr. A population-based case-control study of endometrlal cancer in Shanghai, China. IntJ Cancer 1991; 49: 38-43. 35. Franceschi S, La Vecchia C, Booth M, e# aL Pooled analysis of 3 European case-control studies of ovarian cancer: II. Age at menarche and at menopause. IntJ Cancer 1991; 49: 57-60. 36. Sandier DP. Recent studies in leukemia epidemiolog7. Curt Opin Onco11995; 7: 12-8. 37. Browason RC, Novotay TE, Perry MC. Cigaretxe smoking and adult leukemia. A meta-analysls. Arch Intern Med 1993; 153: 469-75. Cancer Causes and Control Vol 7. 1996