Tobacco Documents Online

Copyright © 1991 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 134, No. 8 Printed in U.S.A. Tobacco Smoke Inhalation Pattern, Tobacco Type, and Bladder Cancer in Spain. - ~,. • . ..~. :.. . ~i ,:i. . " ' ~( :-. :.;. ,: ~ ,?,..... !.: : . .... " .... ., :;..,- ,:' =... ,. • . .. . ., -.:... :. .::.- ...., ~.. • Gonzalo L6pe.z-Abente,1 Carlos A. Gonzhlez,2 Manuel Errezola,3 Antonio Escolar," isabel Izarzugaza,3 Manuel Nebot,s and. E.lio Rib.oil6 bk pa I • .. le.~ ~ po ., Th~ aSSOciati.on b~.tween tobacc.o .sine;king and bladde.r cancer was investigated in a r:nu'l~i(~efiter (~d~e-(~6n~oi, ~tiJ~iy cbndu(~ted in five tbfovJnci~ ~fSl6aln bL=tween: 1983 an.d: . " . " 1986. A match.ed analysis was carried out in males, based on 430 ,histologically confirmed cases, 405 hospital controls, and 386 population controls, matched by age and place of residence. An increased risk was found for smokers as compared with nonsmokers (odds ratio (OR) = 3.79, 95% colafidence interval (CI) 2.41-5.97), and this increase was significantly associated with the intensity of smoking. Smokers of filtei'- tipped cigarettes had a reduced risk as compared with smokers of non-filter-tipped cigarettes (OR = 0.57, 95% CI 0.32-1.02). A diminution of dsk was also observed for smokers of low-tar and low-nicotine ("light") cigarettes. Depth of inhalation was strongly associated with illness. No difference was shown in the logistic regression model between smokers of black tobacco and smokers of blond tobacco after controlling for depth of inhalation. Although the number of persons who smoked blond tobacco exclusively was small, the results suggest that it is important to consider inhalation patterns when studying dsk variations between smokers of black tobacco and smokers of blond tobacco. The age at which a person started to smoke did not appear to affect risk. An analysis of the decrease in risk associated with years since quitting smoking suggested that different components of cigarette smoke may play a role at different stages of the carcinogenic process. Am J Epidemio11991 ;134:830-9. bladder neoplasms; epidemiologic methods; retrospective studie.s; smoking; tobacco The association of cancer of the bladder with tobacco smoking has been demon- strated in numerous studies (1-4). However, them are several aspects 6f smoking wh6se role has not been fully clarified. Questions about the finding of higher risk associated Received for publication June 12, 1989, and in final form March 29, 1991. Abbreviations: CI, confidence interval; OR, odd~ ~atio;. SD, standard deviaf on • 1 Cabc.er Ei3id~miology.Onit, Nati6.na] Center.for ~p.ide- h~iology; .Carlos III l~tl~'tJte o~'Heal~h: Madri~J, Sp~n'. ": ' ": ~ Epidemiology Unit, Hospital Sant Jaume i Santa Mag- rJalena, Matar5, S~ain. ~ Department of Health and Social Security, Basqt~e Government, Vitoria, Spain. 4 Primary Care Department, Andalucian Health Service, C~.diz, Spain. s Municipal Health Institute, Barcelona Town Halt, Bar- celona; Spain. ~ Afialytical Epidemiol(~gy Unit,'lnternational .Acjency for Research on Can6er, Lyon, France. I~epdnt requests .to Dr. Gonzak~ L6pez-Abente, Servicio of Fondo de Investigaci6n Sanitaria (FIS) grant 84/745 and International Agency for Research on Cancer collaborative : research agreement DEB/85/19. .The authors .thank B. Terraccini, F. Berr.ino, and. R. .S~acci for'their hdvice on the..d~isign of the study. They : ~ls.b t'har~l~/~i. Ai~ud~" ar~d'~. Palazl.e'.'.for.:th.eir..Hdp-with " "prepEation of.the data, V..Abraira, F. Pdzo, and M. Posada • for their help and suggestions in interpretation and discus-, sion of the results, and E. Kilbourhe', M. Diez, and I. Villarino for their assistance with the wording of the manuscript. The following oncologists, urologists, and pathologists and institutions, by their collaboration, made this study possible: L Cirera,'M. Gay, and V. Marco (Mutua Terrassa Hospital, Terrassa); J. Fabregat and A. Guzmgua (La Es- p.eranza Hsspital, Barcelona); J." Planas. a.nd A. Gelabert (Hospital .del Mar, Ba~celot~a); A. Badia and J. Castell~. ' be st~ 'i: .... P( a[ of ° hi ff ir tt ir h ~Matar6 HdSlSit~il, Matar6); J. Flor.es and R. Roldan (C&d!z de Epid.emiolosia' del C~,ncer, Centro Nacio.n.al d.e Epi.de: . .Hospital, C.~tdiz); J. Camacho and J. Matz (Jere;; Hospital, • miologia,'Sinesio'Oelgadp, 6, 28029 M~drid, Spain:" • . Cb, diz); E. ~:ahoz and ,h,. Ruiz (L~ U0~a Hos..pital,"C&di.z};" • This .6tudy was" partly ca~rrie.d out:with the finanCial help J. k, lateos" 0.nd. A. EsCud'ero (Ramon and G..a~al H~)spitaJ, '" - . "830 . "2 " " ": - :

Inhalation, Tobacco Type, and Bladder Cancer 831 with use of black tobacco as compared with blond tobacco (5, 6)', the effect of inhalation patterns (7), and other mechanisms of self- regulation controlling blood tar and nicotine cases who could not be interviewed, 103 had died, 39 could not be traced, 32 refused to answer, and 20 could not be included for other reasons. levels (8, .9) have not been completely, re-". : The mean length of time between ~e date .. solved..Moreover,, the re!a.tion betw.een.t,e.m- . of .diagnosis and the interview of the patient IJoral a~p~c~s of sm'okingin eases oF bladder - " was 22"mbnths (siand~/td deviation (SD)6:0) cancer and the phase of the carcinogehic., in prevalent cases and 7 months (SD 3.9) process Upon which smoking acts has not been definitely established (5, 10). Recent studies on the effects of ~arcinogenic com- ponents of smok..~.on, biologic markers (11) and Ofi i~utagenicity of the urine of~ ,rookerY" of different types of.tobacco (12) seem to suggest that smokers of black tobacco are at higher risk of developing bladder cancer than are smokers of blond tobacco. In this analysis, we used data obtained from a multicenter study of bladder cancer in Spain to estimate the risk of this type of tumor in relation to various aspects of smok- ing. MATERIALS AND METHODS The cases were selected from 12 general hospitals spread throughout five Sp~inish provinces (Barcelona, Madrid, C~idiz, Gui- pflzcoa, and Vizcaya). Cases were all male • and female patients with histologically con- firmed carcinoma, polyps, or papilloma of the bladder who were less than 80 years of age on the date of diagnosis and were resid: ing in the province where the hospital was located. All new cases diagnosed between January 1985 and March 1986 (incident cases) and all cases diagnosed between 1983 and 1984 (prevalent cases) were included. Of the 691 pers0ps with bladder cancer di- agnosed ~nd identified in the .hospital regis- ters, 497 wer.e inteffie.wed.(438, males and 59 re.males):. ~mmig.thos'e:i.nt.e .ryiewe.d, 254 (51 perc.e.nt) were iiacident eases, while 243 (49 ~ercent) were prevalent cas~s. Of the 194 Madrid); F. Arocena. R. Telleria, and R. Larburu (Nuestra Se~3ra de Aranzazu Hosp.ital, GuipQzcoa); J. Aurteneche and I. Hemaez (Gu~p~zcoa Hospital, Guip(JzcSa); M. Michelena (The Oncologtcal Institute of GuipQzcoa, GuipQzcoa); A. Perez (Cruces Hospital, Bitb.ao); J. Rores (Bilbao Civil Hospital, Bilbao); X. Elexpe (Department Of Health and Consumer Affairs, Basque Governr0ent, Vitoria); and the Basque Institute of Statistic~ (Vitoria). incident cases: Two controls matched by sex and age (+5 years) were included for each case. One • control.was selected .from the admission reg- " istrr.ofthe .same hospital as.the.ease:. This . hospital control was selected from patients • admitted on the same day or at a later date than the case, and was someone whose dis- charge diagnosis excluded diseases whose risk factors were possibly related to those under study. Patients with the following di- agnoses :were excluded: hematuria, cystitis, urinary lithiasis, diabetes, coronary or cir- culatory illness, digestive or chronic pul- monary illness, and cancer of the respiratory system or upper gastrointestinal system (the oropharynx, esophagus, larynx, and lungs). Of the 827 hospital controls selected, 583 were interviewed. Of 244 who were not in- terviewed, 116 could not be located, 66 had died, 43 refused to answer, and 19 were lost to follow-up for other reasons. The second control was selected at ran- dom from the same section of.the electoral census or municipal register as that of the case. Municipal registers in Spain are contin- ually brought up to date, and the electoral census.is updated every 4 years. Of the 807 population controls selected, 530 were inter- viewed. Of those not interviewed, 142 could not be located, 60 refused to answer, 55 had died, and 20 were lost to follow-up for other .f.eas.on...s. The .ho.spit.al controls i..ncluded 62 wrmen; and ..~he'-p6ptllati+n" controls iia- eluded 65. : The interviews were carried out in the subjects' homes by trained interviewers who were not informed as to the disease and pathology status of those interviewed. The questionnaire contained sections relating to occupational history, tobacco use, passive exposure to tobacco smoke, diet, eonsump- • tion 0.f. coffee, use of artificial gweeteners,

8,32 L6pez-Abente et al. use of analgesics, and past history of diabe- tes, lithiasis, and urinary infections. In the section dealing with tobacco, a complete history was obtained regarding the use of cigarettes, .ici~ars, and pipe tobacco. The in- .t~rciewrr. recorded the:amount, smoked, t.h.e.. ustiai brah~l, the .type afld cli/~r/t.cter~itics Of " the tobacco (black, blond; filtered, hnd/or low-tar and low-nicotine), and the duration of. every period of smoking. Subjects were also asked Whether or not they inhaled the ...... smoke and, if s.o, hq.w deeply: . "':" A randb'm 10 pe}cdht ~ample of all-ques- tionnaires was recoded, by an associate re- searcher who was blind to case-control status in order to assess the level of missing infor- mation and errors in information transcrip- tion. No substantial disagreements were ob- served with respect to the original coding, and there were no statistical differences be- tween cases and controls with respect to frequency of.errors for smoking data. No statistically significant differences were observed among bladder cancer cases, hospital controls, and population controls with respect to socioeconomic status, edu- cational level, number of different jobs held during the person's professional career, and the length of the interview. Smokers were defined as persons who had smoked at least one cigarette per day and/ or one cigar ~ind/or two pipes per week for at least 6 months. Ex-smokers were defined as persons who had ceased to smoke at least 5 years before the date of diagnosis for blad- der cancer cases and 5 years before the date of the interview for controls. In all analyses, jobs found to be high-risk in our data (textile workers, mechanics, adjusted by age and place of residence• The effect estimate was made by conditional lo- gistic regression (14, 15) using a standard program of epidemiglogic analysis (16). For continuous, variables categorized in the a..nalysis,, tests..for trend.were calculated as- sjgnfng/m iflcre'~/sing ifiteger .to each care- ' gbry using ~ regression mrdel.'Mhtchin~ w~as maintained throughout all phases of amil.y~ sis. Analysis was carried out on 430 sets of males (430 cases and 791 controls--405 hos- pital controls and 386 pbpulation controls); .224 questior~a.ires .were-.exclu.ded..bec.a.use the set did not contain at least one 6ase~and one control, and 165 questionnaires from women were excluded because of the scar- city of women smokers (four cases and two controls). In the results presented in tables 2 and 5, we excluded persons who smoked only ci- gars or pipes. The two control groups have been ana- lyzed together, because no difference was found between them with regard to the pro- portions of current smokers, ex-smokers, and nonsmokers (table 1) and because the odds ratios were similar for the average num- ber of cigarettes smoked per day when the two groups were analyzed separately. RESULTS The mean ages of cases, hospital controls, and population controls were 62.4 (SD 11.2), 63.6 (SD 11.0), and 63.1 (SD 11.3) years, respectively. Table 2 shows the relations between the various levels of cigarette smoking and the risk of bladder cancer in males. Age and place of r~sidence were controlled by match- graphic artists, and directors) (13) wer~ t~.st.ed as: pote.ntially cq.nfouu. i.din.g v.arj.'ab.le£...... Lug. For certain associations (cigarettes)day, A tiistoi3r 'of tiriha~y infections v~as '~sb:" • ci~atett~s~iifefl~e, ::~ars 6'f smo~ig,-.and.'. tested as a p.oteritial confofinder. In some of years since" quitting s.moking), .there was a" the analyses presented here (tables 3, 4, 6, and 7), the comparisons were made within the group of smokers because of the need to adjust for other variables of exposure to tobacco. .. A matched analysis was carried out to estimate the odds ratio for each variable. For • that reason, th~ risk ~sfimates presented are. : statistically significant dose response; and the odds ratios in heavy smokers were over 4. With regard to the average number of cigarettes smoked per day, it appears that the relation reaches a plateau above 20 cig- arettes per day. Ho.wever, when one exam- "tries. the. .to.tal nu.rn.b,ers o,f .cigarettes smoked TAB/ 1983 less syn sm sm l sin exl od 16 in Or re te W 1" r( a t: ' t) C ( (

Inhalation, Tobacco Type, and Bladder Cancer 83:3 TABLE 1. Distribution of hospital .and population controls according to their smoking habits, Spain, 1983-1986 Current smokers (cigarettes/day) Cigar Nonsmokers Ex-smokers 1-10 11-20 21-30" >_31 smokers Hospital controls* (n=405) " . . 73. • "" ' "."..- (1.8.2)1" - 15opulatior~ c0ntrols~ (n = 386) 75 (19.6) 89 48 130 38 " " 14 9 • . (22.2) "'.. (12.0) ...(3.2:.4).i'' (9.5)...(3:5).. (2.2,) .... " 107 45 96 40 .. 12 8 (27.9)~ (11.7) (25.!) (10.4) (3.1) "(2.1) - * Data were missing fo( four hospital controls and three population controls. 1 Numbers in parentheses, i~ercentage. "over aiifet[m~i 'tl~e piateau eff~dt l~e~rmes: less evident, probably because this variabl.e synthesizes the information on years of smoking mid average number of cigarettes smoked daily. Risk of bladder cancer decreases with time since stoi~ping smoking, but not to the same extent as in the case of lung cancer (17); the odds ratio for those who gave up smoking 16 or more years prior to interview was 2.37. Risk was similar among incident cases and prevalent cases. We evaluated this in logistic regression models, including an interaction term between the analyzed determinant and an incident or prevalent set indicator. The odds ratio (adjusted for age and prov- ince of residence) for cigar smokers, using only individuals who never smoked ciga- rettes, was 1.32 (95 percent confidence in- terval (CI) 0.41-4.27). Twenty-one persons who smoked cigars exclusivel~r (4 cases and 17 controls) were included in this analysis. Table 3 shows our results for the effects of filter-tipped, non-filter-tipped, and hand- rolled cigarettes. The categories presented are those for smokers of exclusively these types of cigarettes, with the exception of the ."mixed" category, which comprise.s variou~ . "types.of-t.obaccO u.s6. This c.ategory ir~cluded the greatest proportionof ~mokers. an.d; for this reason, .we used it as the refe~:ent when calculating the odds ratios. The results suggest that the filters give a certain amount of protection, independently of the other indicators of exposure. Risk estimates did not vary when. we took into account the number of years of smoking and the date (year) of beginning smoking. The i~rrtlofi. 0f d~ep ihhalers in- ~he group was greater among smokers of filter- tipped cigarettes (59.7 percent) than among smokers of non-filter-tipped cigarettes (28.0 percent). Among smokers of hand-rolled cig- arettes, 41.1 percent inhaled deeply. In the initial analysis including only type of tobacco (unadjusted for number of ciga- rettes or inhalation, and using nonsmokers as the reference group), there was a slightly greater risk for smokers of black tobacco (odds ratio (OR) = 3.71, 95 percent CI 2.39- 5.78) than for smokers of blond tobacco (OR = 3.16, 95 percent CI 1.52-6.58). Table 4 shows results for type of tobacco (black vs. blond) after adjustment for num- ber of cigarettes smoked per day, depth of inhalation, and years of smoking, using smokers of black tobacco as the reference group. We found no difference in this analy- sis between the two types of tobacco. The categories included those who smoked one or the other of the two types of tobacco exclusively• The confounding effect of the~ variable "depth of inhalation" was higher than that of the covariate "number of ciga- rettes smoked per day." The proportion of inha.l~rs was greater amon~ users of black tob.ac.c9 tho/a, among users of, .blond t0ba.c~b, "; • - for brth dase.s and controls (90" perdent vs. • 84 •percent in "cases and 79 percent vs. 65 " percent in controls). We do not know what level of precision was achieved in the gathering of information on inhalation. We assumed, in this analysis, that the smoker knew this aspect of his smoking well, although the depth of his in- halation may have v.aried over his lifetime.

834 L6pez-Abente et al. TABLE 2. Odds ratios for bladder cancer in males according to degree of cigarette smoking (as compared with nonsmokers), Spain, 1983-1986 • Variabte and categon/ , No. of No. of • OR*.'{" 95% . . cases controls .' • Smoking :. .: .'-, .:Nonsmoker • .. : EYer'~/noker " Nonsmoker " Ex-smoker. Curre, nt s~oker 27 ~7 .90 309 148 1.00 196 2.69 426 4.37 Ever smoking(average no. ofcigare~es/day) 0 27 , 1-10 59, 11-20 "':" '" ": " " "": .245 :~'. " ." 21-30 66 >30 26 148 174 " 305 " .'... 102 " "" 37 (Referent).. 2.4i -5.97 • "(Referent) : 1.61 -.4.49. 2.75-:,6.95 1.00 (Referent) 1.88 1.10-3.20 .4.84.. , . 3;01-7.78 • 4.10 2.3~-~.13 4.21 2.11-8.43 X2 trend~: = 40.0 p < 0.0001 Current smoking (average no. of cigarettes/day) 0 27 148 1.00 (Referent) 1-10 39 93 2.40 1.35-4.28 11-20 201 226 5.28 3.25-8.57 21-30 48 78 3,92 2.17-7.07 >30 19 26 4.34 2.04-9.24 x~ trend.l: = 31.5 p < 0.0001 Years of smoking 0 27 148 1.00 (ReferenO 1-19 " 23 66 1.80 0.94-3.45 20-39 137 221 3.91 2.35-6.49 40-59 222 311 4.66 2.87-7.54 ->60 17 24 4.44 2.02-9.78 x=trend~ = 41.5 p < 0.0001 Years since quitting smoking Nonsmoker 27 148 1.00 (Referent) ->-16 38 91 2.37 1.32--4.26 6-15 52 105 2.99 1.71-5.22 0-5 309 426 4,38 2.75-6.97 X2 trend~c = 44.0 p < 0.0001 Ufetime smoking (no. of cigarettes/lifetime) 0 27 148 1.00 (Referent) <150,000 73 195 2.21 1.33-3,68 150,000-299,999 • .. 183 242 4.56 2.82-7.37 ->300,.000.... 140 181 ,4.88 2.94-8.08 • • .::. ",-~ . ,'." ;-. ~:i;.~.~'tr.end:l:.~'~,9.9,. • • -~ _ p < 0.0001 Mi ". :N~ . Fil i smc sine ~. . ~ . ana 1 * OR. odds ratio; CI, confidence ~nterv~L " " 1" Adjusted for age and place of residence (matched conditional Iogis{ic regression analysis). Odds ratioswere "dedved from the matched analysis and cannot be calculated directly from the unmatched distribution of cases and control~ shown in the table. ~ Trend test for dose response. However, the results in table 5 appear to be particularly cons!stent, since one may ob- serve a clear variation of effect based on depth of inhalation.- .. The inhalation covariate, in theoretic terms~ is a modifier of the effect of other aspects of smoking, That is, the effect asso- ciated with the number of cigarettes smoked .

Inhalation, Tobacco Type, and Bladder Cancer 835 TABLE 3. Odds ratios for bladder i:ancer in male cigarette smokers a,ccording to use of filter-tipped cigarettes, Spain, 1983-1986 Type of cigarette No. of No. of OR*, i 95% CI* cases controls Mixed use 330 469 .1..00 (Referent) .. " Hand-.rolled 30 " 58 " 0.9"4 .0.54z1.66 • Filter-tipped .26 69 0.57 .0.32-1.02 • * OR, odds ratio; CI, confidence interval. 1 Adjusted for age, place of residence, no. Of cigarettes smoked per day, depth of inhalation, cigar smoking, and years since quitting smoking (matched conditional logistic regression analysis). per day varies betWeen inhalers and nonin-. balers. In our analyses, there was a signifi- cant interaction between number of ciga- rettes smoked per day and depth of inhala- tion. The. result of the likelihood ratio test for the model with and without an interac- tion term was x2 (6 df) = 14.90 (p = 0.02). In table 6, we present the estimates of risk for both, taking into account the interaction term. Among persons who inhaled deeply, number of cigarettes smoked per day was a less important determinant of bladder can- cer risk. Low-tar and low-nicotine ("light") ciga- rettes have only recently been put on the market in Spain. The number of smokers who smoked this type of tobacco exclusively was very small (nine cases and one control). In table 7, we grouped together thos~ who smoked exclusively light tobacco and those who smoked both light and regular tobacco and compared them with those who smoked only regular tobacco. A diminution of risk was observed from the smoking of some light tobacco. The tar and nicotine content of t~)bacc0 has gone down in recent, decades (18). We. • could -riot .detect differences.in-., odds ratios by yea~: of starting to smoke. For those Siab- jects who started smoking in 1955 or earlier (using as the reference group subjects who started after 1955), the odds ratio was 1.07 (95 percent CI 0.50-2.30). The estimate was adjusted for age; place of residence, number of cigarettes smokedper day, depth of in- .halation, and years since quitting smoking. The age at which a person started to smoke did not appear to affect the risk. Having a high-risk job was not a con- founding variable in any of the analyses. The odds ratio for a person who smoked more than 30 cigarettes per day and worked in a high-risk job was 9.6. The odds ratio for having a high-risk job was 2.03 (95 percent CI 1.49-2.77) (13). Assuming the suitability of the logistic regression model in this analy- sis, these data suggest the existence of a multiplicative effect for the two exposures. in figure 1, we show odds ratios for ex- smokers by time since cessation of smoking (using current smokers as the reference cat- egory) and by the presence or absence of occupational exposure. The figure shows a rapid decrease in risk in the years that follow quitting smoking but later on a slight in- crease and stabilization. In persons with TABLE 4. Odds ratios for bladder cancer in male cigarette smokers according to type of tobacco smoked, Spain, 1983-1986 Black tobacco] Blond tobacco Mixed tobacco Potential confounders (n = 286/461):1: (n = 19/34) . (n = 93/•128) considered* .... ' OR§. -OR " 95?/o el§ " .. OR • "95% Cl '. None 1.00 0.99 0.522i.88 " ~.08 0.78-1.50 No. of cigarettes/day 1.00 1.1 4 0.59-2.18 1.05 0.76-1.47 Depth of inhalation 1.00 1.25 0.65-2.43 1.07 0.76-1.50 No. of cigarettes/day, depth of inhalation, and years of smoking 1.00 1.40 0.71-2.75 0.99 0.70-1.39 * Variables adjusted for in addition to matched variables (conditional logistic regression analysis). 1 Referent. :1: No. of cases/no, of controls. § OR, odds ratio; CI, confidence interval.

836 L6pez-Abente et al. TABLE 5. Odds ratios for bladder cancer in male cigarette smoke~s according to depth of inhalation, Spain, 1983-1986 No. of No. of cases ~ntrols OR*, 1 95% CI* Nbnsmoker. " 27 14~ .1.00; (Referent) No inhalation 35 124". 1.5.0" 0.84~-2.69 ' Depth of inhalation Moderate 87 143 3.67 2.16-6.24 Deep 272 352 4.86 3.02-7.82 * OR, odds ratio; CI; confidence interval. 1" Adjusted for ag.e and p!ace of residence (matched condi- tional logistic regression analy.sis). TABLE 6. Odds ratios* for bladder cancer in m.ale cigarette smokers according to number of cigarettes smoked per day and depth of inhalation, Spain, 1983-1986 No. of No Moderate Deep cigarettes/day inhalation inhalation inhalation 1-10 1.001" 2.05 1.91 11-20 1.62 3.89 4.59 21-30 2.62 3.18 3.92 >30 4.24 3.56 • 3.84 * Adjusted for age, place of residence, and use of filter- tipped cigarettes (matched conditional logistic regression analy- sis). 1" Referent. TABLE 7. Odds ratios for bladder cancer in male cigarette smokers according to use of low-tar and low-nicotine ("light") cigarettes, Spain, 1983-1986 No. of No. of OR*, 1" 95% cases controls Regular ciga- rettes 380 572 1.00 (Referent) Both light and regular ciga- rettes 16 48 0.49 0.25-0.96 * OR, odds ratio; CI, confidence interval 1" Adjusted for age, place or residence, use of filter-tipped cigarettes, no. of cigarettes smoked per day, and inhalation (matched conditional logistic regressi .on analysis). ~gh-~isk j6b~ (oCC~l~atibnalls; exposed), the'" pattern of the curve is different. The population attributable risk (19) as- ' sociated with cigarette smoking was 67 per- cent. DISCUSSION The adjusted mortality rate for bladder cancer among men in Spain increased 60 percent between 1955 and 1975 (20). In )DD8 RATIO 1.6" 1.4 ..... =~¢_ ~ -~- NOT OGGUPAT. EXP. 0.8" "1 0 10 20 80 40 50 YEARS SINCE QurVrlNG SMOKING . " F~uB'E'I. 06d~ ratios~for'bladder'eit~cer ln'~naie-e-~- : " " smokers versus male current smokers, by years since quitting smoking and by occupational exposure, Spain, .1983-.1986. Estimates were adlusted for liletime num- ber of cigarettes smoked. Age and place of residence were controlled by matching. 1985, the adjusted mortality rate (world standard population) in women was 1.10 per 100,000 person-years, as compared with 7.81 per 100,000 person-years in men (un- published data). Smoking was extremely rare among the women in this study; their average age was 68 years. Tobacco smoke contains a large number of carcinogens. Two of them are 2- naphthylamine and 4-aminobiphenyl, both known to cause bladder cancer (11, 21). The relation between dose and response shown in table 2, based on the average num- ber of cigarettes smoked per day, shows a leveling off and even a negative slope for smokers of more than 20 cigarettes per day. This leveling of the odds ratios has been described by others (4, 5, 22, 23). A number of different explanations have been sug- gested, including an underestimation of to- bacco use among heavy' z.mokers (22), .the ~i~tence of diseases.reJ.atedtd-s'mqldng that appear .~arlier than" bIa.dde'r cancer (4);" ~nd imperfect comparability between dases and controls because of design problems (5). This leveling off lessens when another variable such as total or lifetime smoking is analyzed. However, this stabilization of risk has not been observed for other tumors of the uro- logic system such as those of the renal pelvis and ureter (24). The stabilization of the mutagenic effect of urine observed by Malaveille et al. (12) ir 20 or more cigarettes I ing. Recent studies ha)e •. of a ~reater risk associ )~obac~o cigarettes, ~ twice the risk as d6 blt (5, 6, 25), arid it has b greater concentration s~cific to bladder c~ black tobacco .might e .otrr.s.t.udy~ howeyer, t} seen. The relative risk Italy was higher than United States and En~ that could be due to black tobacco smoki countries. The habit bacco exclusively was uiation we studied. Other hypotheses w. difference in findings ! others are errors in th, of tobacco and an ac: inhalation as a risk fa etiology. As for the methoc interviewed with rega used (black or blond), a list of all the type Spain. The interview, instructions on how taught to consult it. Ifa error, it would have be would have masked ~ blond and black toba~ In the logistic regre we evaluated the diffe blond and black tobac seemed to be a ne~ati' S'..moki~g- p .atterns ..di who smbked exclusiv those who smoked bacco. Smokers of t more (and smoked ~r vs. 12.4 cigarettes/da: than smokers of Non, ~Depth of inhalafio in the estimation of . other case-control stu ated with depth of int

Inhalation, Tobacco Type, and Bladder Cancer 837 Malaveille et al. (12) in persons who smoked 20 or more cigarettes per day is rather strik- ing. Recent studies have revealed the existence of a greater risk associated with Use of black tobacco cigarettes, whidh present a]most twice the risk as do blond tobacco cigarettes (5, 6, 25), and it has been suggested that the greater concentration of aromatic amines specific to bladder cancer in the air-cured black tobacco might explain this finding. In our ~tudy, howe'~er, this difference wa~ not. seen. The relative risk observed in Spain and Italy was higher than that observed in the United States and England (4, 26), a finding that could be due to the predominance of black tobacco smoking in Mediterranean countries. The habit of smoking blond to- bacco exclusively was infrequent in the pop- ulation we studied. Other hypotheses which might explain the difference in findings between our study and others are errors in the classification of type of tobacco and an actual effect of depth of inhalation as a risk factor in bladder cancer etiology. As for the method of classifying those interviewed with regard to type of tobacco used (black or blond), coding was done from a list of all the types of tobacco used in Spain. The interviewers had this list, .with instructions on how to use it, and were taught to consult it. If there was classification error, it would have been nondifferential and would have masked the difference between blond and black tobacco (27). In the logistic regression model in which we evaluated the differences in risk between blond and black tobacco, depth of inhalation see.med .to .be a negative ct)nfounding factor. Smoking 15atterns differed between those who smoked exclusively black tobacco and those who smoked exclusively blond to- bacco. Smokers of black tobacco inhaled more (and smoked more: 16 cigarettes/day vs. 12.4 cigarettes/day in the control group) than smokers of blond tobacco. Depth of inhalation was very important in the estimation of risk in our study. In other ease-control studies, no effects associ- ated with depth of inhalation were noted (3, 26), although, in one of them (26), the ques- tion was posed differently. The finding of an effect associated with depth of inhalation is plausible, since it is pr6sently accepted that the intake of various components of tobacco depends not only ori th~ /iumber of ciga- rettes smoked biat also on the extent and frequency of puffing and the depth of inha- lation (8, 9, 28-30). A change in the type of tobacco used produces, an adjustment of smoking habits which tends to ~lf-regulate blood tar and. nicotine levels (8, 9). Therefore, the assess- ment of the separate effects of blond and black tobacco could be improved, avoiding misclassification, by analyzing clean cate- gories of exposure (i.e., by analyzing persons who smoked exclusively one type of cigarette or the other). Malaveille et al. (12), in their study of levels of mutagens in the urine of smokers of black and blond tobacco, found twice as much mutagenic effect in smokers of black tobacco as in smokers of blond tobacco when they adjusted for nicotine and cotinine levels in urine. According to their results, the condensate of black tobacco central smoke contains more nicotine than that of blond tobacco smoke, but per cigarette smoked, smokers of black tobacco excrete less nicotine in the urine; this is why their findings on mutagenicity were confounded by the number of cigarettes smoked. Indeed, when calculating the mutagenic effect ac- cording to number of cigarettes smoked per day, Malaveille et al. could not find differ- ences between the two types of tobacco (12). Although persons who smoke black to- bacco present higher levels of hemoglobin adducts of aromatic amines than those who smoke blond tobacco (adjusting. for the number of cigarettes), 4-aminobiphenyl and 3-aminobiphenyl levels are equal in the cen- tral smoke of both types of tobacco (11). We do not know how inhalation pattern might have affected these results. We did not find risk variations in relation to age at starting smoking. This is in agree- ment with previous observations (5). In figure 1, there is a rapid decrease in risk in the years that folIow quitting smok-

838 L6pez-Abente et al. ing, but in later years a slight increase and stabilization can be observed. Ttae first part of the curve is .compatible with a late-stage • .carcinogenic" mecha.nism and the second with an early-stage mechanism (31). Quit- ting smoking does not result in the risk's returning to a nonsmoker's level, probably because the carcinogenic biologic mecha- nism has already been triggered. It is possible that cigarette smoke .has some effect'on both ph~ises of the carcinogeriic process in bladder cancer, as happens in lung.cancer (32). In persons who are occupationally exposed to chemical carcinogens, the pattern of the curve is different. Although temporary as- pects of job exposure have not been con- trolled for in this analysis, this pattern might represent an effect modification phenome- non. The results shown confirm that the risk of bladder cancer's developing in cigarette smokers depends on the intensity of the habit, the number of years of smoking, and the amount of tobacco smoked during a lifetime. We think that inhalation intensity (fre- quency and depth) should be measured at each different point in the tobacco use history of an individual. Since it might be difficult to conduct these measurements, laboratory studies should be carried out to measure exposure by means of biologic markers for inhalation patterns and for type and amount of tobacco consumed. REFERENCES 1. Lilienfeld AM,.Levin ML, Moore GE. The associ- ation of smoking" with cancer of the urinary bladder in humans. Arch Intern Med 1956;98:129-35. 2. Cole P, Monson RR, Haning H, et al. Smoking and cancer of the lower urinary tract. N Engl J Med 1971;284:129-34. 3. Howe GR, Burch JD, Miller AB, et al. Tobacco use, occupation, coffee, various nutrients, and blad- der cancer. J Natt Cancer Inst 1980;64:701 - 13. 4. Cartwright RA, Adib R, Appleyard I, et al. Ciga- rette smoking and bladder cancer: an epidemiolog- ical inquiry in West Yorkshire. J Epidemiol Com- munity Health 1983;37:256-63. 5. Vineis P, Esteve J, Terracini B. Bladder cancer and smoking in males~ types of cigarettes, age at start, effect of stopping, and interaction with occupation. Int J Cancer 1984;34:165-70. 6. Iscovich J, Castelletto R, Esteve J, et al. Tobacco smoking, occupational exposure, and bladder can- cer in Argentina. Int J Cancer 1987;40:734-40. 7. Wald NJ, Idle M, Boreham J, et al. Inhaling habits among, smokers of different types of cigarette. Thorax 1980;35:925-8. 8. Herning RI, Jones RT, Benowitz NL, et al. How a cigarette is smoked determines blood nicotine lev- els. Clin Pharmacol Ther 1983;33:84-90. 9. Sutton SR, Russell MAH, Iyer R, et al. Relation- ship between cigarette yields, puffing.patterns, and smoke intake: evidence for tar compensation? Br MedJ 1982;285:600-3.. ., 10. Decarli A, Peto J, Piolatto G, et al. Bladder c~ncer" moi-tality of workers exposed to aromatic amines: analysis of models of carcinogenesis. Br J Cancer 1985;51:707-12. 11. Bryant MS, Vineis P, Skipper PL, et al. Hemoglo- bin adducts of aromatic amines: associations with smoking status and type of tobacco. Proc Natl Acad Sci USA 1988;85:9788-91. 12. Malaveille C, Vineis P, EstEve J, et al. Levels of mutagens in the urine of smokers of black and blond tobacco correlate with their risk of bladder cancer. Carcinogenesis 1989; 10:577-86. 13. Gonz~ez CA, L6pez-Abente G, Errezola M, et al. Occupation and bladder cancer in Spain: a multi- centre case-control study. Int J Epidemiol 1989; 18:569-77. 14. Breslow NE, Day NE, Halvorsen KT, et al. Esti- mation of multiple relative risk functions in matched case-control studies. Am J Epidemiol 1978;108:299-307. 15. Breslow NE, Day NE, eds. Statistical methods in cancer research. Vol I. The analysis of case-control studies. Lyon, France: International Agency for Research on Cancer, 1980. (IARC scientific publi- cation no. 32). 16. EPILOG. Version 3.06. Pasadena, CA: Epicenter Software, 1986. 17. Benhamou S, Benhamou E, Tirmarche M, et al. Lung cancer and use of cigarettes: a French case- control study. J Natl Cancer Inst 1985;74:1169-75. 18. Klus H. Low-tar cigarettes: possibilities and limi- tations. In: Zaridze D, Peto R, eds. Tobacco: a major international health hazard. Lyon, France: International Agency for Research on Cancer, 1986:265-81..(IARC scientific publication no. 74). 19. Miettinen OS. Proportion of disease caused or pre- vented by a given exposure, trait or intervention. Am J Epidemiol 1974;99:325-32. . ' " . 20. L6pez-Abente G: Bladfder cance/"in Spain: mortal" ity trends (i 955-1975). Cancer 1983;51:2367-70. 2 I. Hoffmann D, Wynder EL. Chemical constituents and bioactivity of tobacco smoke. In: Zaridze D, - Peto R, eds. Tobacco: a major international health hazard. Lyon, France: International Agency for Research on Cancer, 1986:145-65. (IARC scien- tific publication no. 74). 22. Wynder EL, Goldsmith R. The epidemiology of bladder cancer:, a second look. Cancer I977;40: 1246-68. 23. Schifflers E, Jamart J, Renard V. Tobacco and occupation as risk factors in bladder cancer: a caser control study in southern Belgium. Int J Cancer 1987;39:287-92. " 24. Jensen OM, Knudsen JB, McLaughlin JK, et al. The C, and u~ tional. 25. Vineis and ty • cancer • 26. Hartgc cigaret contro 27. Copek Bias d relatiw 28. Peach ..... inexpe Peto R hazard

Inhalation, Tobacco Type, and Bladder Cancer 839 27. The Cope~hhgen case-control study of renal pelvis and ureter cancer: role of smoking and occupa- tional exposures. Int J Cancer 1988;41:557-61. 25. Vineis P, Esteve J, Hartge P, et al. Effects of timing and type of tobacco in cigarette-induced bladder cancer. Cancer Res 1988;48:3849-52. 26. Hartge P, Silverman D, Hoover R, et al. Changing cigarette habits and bladder cancer risk: a case- control study. J Natl Cancer Inst 1987;78:1119-25. Copeland KT, Checkoway H, McMichael A J, et al. Bias due to misclassification in th6 estimation of relative ri~k. Am J Epidemiol 1977;105:488-95. 28. Peach H, Morris RW, Ellard GA, et al. A simple, inexpensive t/rine test of smoking. In.: Zaridze D, Peto R, eds. Tobacco: a major int.emational health hazard. Lyon, France: International Agency for 29. Research on Cancer, 1986:183-93. (IARC scien- tific publication no. 74). Vogt TM, Selvin S, Hulley SB. Comparison of biochemical and questionnaire estimates of tobacco exposure. Prev Med 1979;8:23-33. 30. Slattery ML, Schumacher MC, West DW, et al. Smoking and bladder cancer: the modifying effect of cigarettes on other factors. Cancer 1988;61: 402-8. 31. Day NE, Brown CC. Multistage models and pri- mary prevention of cancer. J Natl Cancer Inst 1980;64:977-89. 32. Brown CC, Chu K. Use of multistage models to infer stage affected by carcinogenic exposu.re: ex- ample of lung cancer and cigarette smoking. J Chronic Dis 1987;40(suppl 2):171S-9S.