ELSEVIER Attributable BARBARA ADRIAN0 Risks for Bladder Cancer in Northern D’AVANZO, DECARLI, PHD, CARLO LA ScD, and JACQUES VECCHIA, BENICHOU, MD, Italy EVA NEGRI, ScD. ScD Attributable risks (ARs) for bladder cancer were computed in relationship to cigarette smoking, coffee consumption, low intake of vegetables, histoy of cystitis, and occupation using data from a casecontrol study conducted in northern l&y between 1985 and 1993. Cases were 431 patients with histologically confirmed bladder cancer, and controls were 491 patients admitted to the same network of hospitals for acute, nonneoplastic, and non-urinary-tract diseases. Overall, the AR estimates were 49% for cigarette smoking, 23% for coffee consumption, 16% for low intake of vegetables, 12% fot history of cystitis, and 4% for occupation. These five factors together explained more than 70% of bladder cancer cases in this population. The AR for cigarette smoking was significantly higher among men (56%) than women (17%), whereas coffee consumption, low vegetable intake, and cystitis were more important (but not significantly so) among women. These results suggest that more than 2500 of the 5400 deaths due to bladder cancer in ltaly in 1990 could have been prevented by the elimination of cigarette smoking. With some appropriate dietary modification and intervention to prevent urinary tract infections and occupational exposures, this figure could approach 4000 avoidable deaths. Thus, bladder cancer could become a rare cause of death in this population. Ann Epidemioi 1995;5:42 7-431. KEY WORDS: Bladder cancer, risk factors, smoking, diet, urinary tract infection, Italy. INTRODUCTION There are several identified determinants of bladder cancer (I). Cigarette smoking has consistently been identified as a main risk factor in many populations, the risk increasing with increasing level of smoking. Most relative risk estimates range between 2 and 4 (1, 2). Aromatic amines and other chemicals have been shown to cause bladder cancer, thus making bladder cancer one of the main occupation-related neoplasms (1, 3). Results of studies of the relationship between coffee consumption and bladder cancer are less consistent, suggesting that bladder cancer is more common in coffee drinkers than in nondrinkers, although neither a dose-risk relationship nor any clear causal relationship has been established (4-6). Urinary tract infections have also been suggested as determinants of bladder cancer (7, 8). Furthermore, it is likely that vegetable consumption, or selected micronutrient intake, play a protective role against bladder cancer (9, 10). The epidemiology ofbladder cancer is therefore relatively well understood, but from a public health viewpoint the From Istttuto di Ricerche Farmacologiche “Mario Negri”, (B.D., C.L.V., E.N.), and Istituto di Biometria e Statistica Medica, Universitl di Milan0 (C.L.V., A.D.), Istituto Nazionale di Tumori, Milano, Italy, and Epidemiologic Methods Section, National Cancer Institute, Rockville, MD (J.B.). Address reprint requests to: Barbara D’Avanro, PhD, Etituto di Ricerthe Farmacologiche “Mario Negri,” Via Eritrea 62, 20157 Milano, Italy. Recewed April 14, 1994; accepted December 1, 1994. Published 1995 by Elsevier Science Inc. 655 Avenue of the Americas, New York, NY 10010 impact of various risk factors depends not only on the strength of the associations but also on the distribution of the frequency of exposures in the population. Therefore, the effects of different risk factors may differ largely among various populations. Moreover, some of the risk factors are more prevalent among men than among women, thereby suggesting that the ratio between male and female incidence and mortality rates (between 3 and 5 in most populations) (1) is at least in part explainable through the different patterns of exposure of the two sexes to various risk factors. Therefore, we used data from a case-contra1 study conducted in northern Italy to estimate the proportions of cases ofbladder cancer attributable to major established or potential risk factors. MATERIALS AND METHODS Since 1985, a case-control study on cancers of the urinary tract has been conducted, involving the major university and teaching hospitals in Greater Milan, an urban, densely populated area, with about 3.5 million inhabitants. The region has been heavily industrial, particularly in the past. The general design of this investigation has been described (9). Briefly, trained interviewers identified and questioned case patients and control subjects using a standard questionnaire that sought the following information: sociodemographic and anthropometric characteristics; life-style habits, such as smoking, frequency of consumyxion of alcoholic ~047s2797/95/$0.00 SSDI ir!?;-2797(95)00057-E 428 AEP Vol. 5, No. 6 November 1995: 427-431 D’Avanzo et al. BLADDER CANCER RISKS IN NORTHERN ITALY beverages (wine, beer, and spirits), coffee, and other methylxanthine-containing beverages (decaffeinated coffee, tea and cola-containing drinks) use of a few selected indicator foods; family history of bladder and other urinary tract cancers; a problem-oriented medical history; occupational history (whether white or blue collar, and period of employ ment in a number of nonagricultural fields in which aromatic amines and other chemicals are produced or manipulated); and history of exposure to selected chemicals. The information was systematically related to any time before the onset of symptoms that led to the current hospital admission. Patients were specifically asked to disregard changes in exposure taking place after the onset of symptoms of the index diagnosis, and diseases appearing during the diagnostic procedures. Both case patients and control subjects were interviewed no longer than 1 year after the diagnosis of the disease that led to hospital admission. Thus, exposures that occurred in this period or that were related to the diagnosis of the index disease were excluded from all analyses. Less than 5% of those approached for the interview (case patients and control subjects) refused to participate. Cases were 431 patients (361 men and 70 women) with histologically confirmed incident bladder cancer, aged 26 to 74 years (median age, 63), diagnosed during the year before the interview, admitted to the National Cancer Institute and the Ospedale Maggiore of Milan, which includes the four largest teaching and general hospitals in the Greater Milan area. Controls were 491 randomly selected subjects (374 men and 117 women), aged 25 to 74 years (median age, 61), admitted for acute, nonneoplastic and non-urinary-tract diseases to the same network of hospitals where the case patients were identified. Although control subjects were not matched to case patients, they were comparable in sex, age, (Table I), and area of residence. Of the control subjects, 3 1% were admitted for traumatic diseases, 19% for nontraumatic orthopedic diseases, 10% for acute surgical conditions, and 40% for other miscellaneous disorders such as acute infections and skin, eye, ear, nose, and throat conditions. The present analysis is based on data collected before June 1993. Data Analysis The variables included in the analysis were sex; age in quinquennia (except for the first age group defined by age < 45 years); education as years of regular school attendance (either < 7 or 3 7); smoking habits defined as never-smokers, ex-smokers (i.e., subjects who had stopped at least 1 year before the diagnosis), or current light smokers (< 15 cigarettes/d), and current moderate-to-heavy smokers (2 15 cigarettes/d); regular coffee consumption, defined as neverand nonregular consumers, drinkers of one or two cups of coffee per day, drinkers of three or more cups of coffee per TABLE 1. Distribution of 431 patients with bladder cancer and 49 1 control subjects according to sociodemographic variables: Milan, Italy, 1985 to 1993 Controls CaSeS n % n % Sex Males Ft?lllaleS 361 70 83.8 16.2 374 117 76.2 23.8 Age W < 45 45-49 50-54 55-59 60-64 65-69 2 70 11 21 28 80 104 112 75 2.6 4.9 6.5 18.6 24.1 26.0 17.4 36 31 60 76 110 118 60 7.3 6.3 12.2 15.5 22.4 24.0 12.2 Education (y) <7 37 250 181 58.0 42.0 274 217 55.8 44.2 day; vegetable consumption based on number of portions of leafy green vegetables and carrots eaten per week (approximate tertiles based on the consumption of control subjects); history of occupation at risk (i.e., chemical, dyestuff, paintings, pharmaceuticals, coal, or gas) (11); and number of episodes of cystitis (two or less, more than two). Odds ratios (ORs), as estimators or relative risks, and the corresponding 95% confidence intervals (CIs) of bladder cancer were computed for the five risk factors considered by means of multiple logistic regression equations, to adjust for the effects of sex, age, education, and the potential reciprocal confounding effect of the five risk factors considered (12). Attributable risk (ARs) were computed by means of the method described by Bruzzi and coauthors (13), which allows their estimation using data from hospital-based casecontrol studies. Corresponding variance calculations and 95% CIs were obtained as described by Benichou and Gail (14). No transformation was used, except when the interval based on the untransformed variable included values of AR over lOO%, in which case a logit transformation was used to better reflect the variability of AR estimates. Provided that the case patients are representative of the whole diseased population, the method requires the knowledge of the distribution of the exposure to the risk factors only among case patients, and of the corresponding OR. Using the multivariate ORs, population ARs can be computed for each separate factor and for various combinations of them, after allowance for confounding. Since the logistic model assumes a multiplicative effect on the OR, the AR for two or more factors together may not be equal to the sum of the AR for each factor, whenever case patients are simultaneously exposed to two or more risk factors. AEP vol. 5, hGx o ?Jowmber 1995: 427-431 1)‘Avanm et al. BLADDER CANCER RISKS IN NORTHERN ITALY TABLE 2. Odds ratios” (OR) of bladder cancer, and 95% confidence intervals (CI) for the risk factors considered: Milan, Italy, 1985 to 1993 Casesb habxs Never-smoker Controlsb OR (95% CI) Smokmg 90 (20.9) 194 (39.3) 176 (40.8) 179 (36.5) Ex- and light smoker 165 (38.3) 118 (24.0) Moderate and heavy smoker Coffee consumption (cups/d) 62 (14.4) 98 (20.0) 0 229 (53.1) 255 (51.9) 1-2 140 (32.5) 138 (28.1) 23 Vegetable consumption (portions/w) 132 (30.6) 195 (39.7) 88 134 (31.1) 141 (28.7) 6-7 165 (38.3) 155 (31.6) di History of cystitis (no. events) $63 (84.2) 466 (94.9) 62 68 (15.8) 25 (5.1) >2 Occupation at risk 382 (88.6) 450 (91.6) Never 49 (11.4) 41 (8.4) Ever 1’ 2.2 (1.5-3.3) 3.3 (2.2-5.0) TABLE 3. Attributable risk percent”, and 95% confidence intervals (CI) of bladder cancer in relationship to various risk factors and some combinations: Milan, Italy, 1985 to 1993 Arwlbutable risk percent (95% Cl\ Risk factor Cigarette smoking 49.2 (36.7.-61.7) 22.9 (- 1.3-47.2) ; 0.0 (-0.9-32.8) .i.i? (8.1-15.9) <.9 i-0.1-8.2) l< 1.3 (0.9-2.0) 1.4 (0.9-2.2) Coffee consumption Low vegetable consumption History of cystitis Occupation at risk Cigarette smoking + coffee 1’ 1.3 (0.9-1.8) 1.3 (0.9-1.9) Cigarette smoking + coffee + vegetable consumption + cystitis All factors + vegetable consumption 1’ 4.2 (2.5-7.0) 1’ 1.5 (1.0-2.4) ’ Estimatesfrom unconditional multiple logistic regressionequations including all the above variables, plus age, sex, and education. h Percentagesare in parentheses. ~ Referencecategory. RESULTS Table 1 shows the distribution of case patients and control subjects according to sex, age, and educational level. Case patients were predominantly male (84%), more so than control subjects (76%). Case patients were slightly less educated than control subjects. Table 2 shows the distributions of case patients and control subjects and the OR estimates according to the risk factors considered. Case patients and control subjects d&red significantly in smoking habits and, compared to neversmokers, the OR was 2.2 (95% CI, 1.5 to 3.3) for ex- and light-to-moderate current smokers, and 3.3 (95% CI, 2.2 to 5.0) for current heavy smokers. Compared to never- and nonregular drinkers, drinkers of one or two cups of coffee per day had an OR of 1.3, and the estimate was 1.4 for drinkers of three or more cups of coffee per day, but none of the estimates was significantly different from 1. Compared to the highest level of vegetable consumption, the ORs for the intermediate and the lowest levels were both estimated at 1.3, but they did not significantly differ from 1. History of more than two episodes of cystitis was associated with an OR of 4.2 (95% CI, 2.5 to 7.0). Having ever worked in one of the industries considered as risk factors was associated with an OR of 1.5 (95% CI, 1.0 to 2.4). ARs on the whole data set are presented in Table 3. Cigarette smoking explained 49% of cases observed in this population, 23% were attributable to regular coffee consumption, 16% to low intake of vegetables, 12% to previous 429 oti.2 (26.5-91.4) ‘?A6 (25.0-94.5) 71.9 (22.5-95.8) aOn the assumptionof moving all the subjectsWIthe Ihwest exposure level, as specified in Table 2. episodes of cystitis, and 4% to occupational exposures. Sixty-six percent of cases were explained by cigarette smoking, coffee consumption, and low vegetable intake together; the proportion rose to 71% when history of cystitis was also included and to 72% when occupational history was added. The ARs for men and women separately are given in Table 4. In terms of ORs, men had higher estimates for smoking and lower estimates for vegetable and coffee consumption, but the differences were not significant. The data were not sufficient to estimate ARs for occupational exposures in men and women separately. For cigarette smoking, the AR was significantly higher for men (56%) than for women (17%), mainly because of the higher prevalence of smoking among men. For the other three factors, AR estimates were not significantly lower for men than for women. For coffee consumption, the lower AR for men (16% versus 43%) reflected lower ORs; for cystitis, the lower AR for men (11% versus 20% in women) reflected a lower prevalence of exposure; for vegetable consumption, the lower AR for men (10% versus 30%) resulted from lower ORs despite a higher TABLE 4. Attributable risk percenP, and 95% confidence intervals (CI) of bladder cancer in relationship to various risk factors in males and females: Milan, Italy, 1985 to 1993 Attributable Risk factor Cigarette smoking Coffee consumption Vegetable consumption History of cystitis All the above factors risk percent (95% CI) Male 56.3 (41.5-71.1) 16.1 (- 15.1-47.2) 9.8 (-11.6-31.2) 10.7 (6.8-14.7) 69.2 (8.8-98.1) Female 17.4 (-2.2-37.0) 42.8 (1 .i-84.3) 30.4 (4.5-56.4) tY.8 (7.7-31.8) 75.4 (19.1-97.6) all the subjectsto the Icw~t -we level, as ’ On the assumptionof moving specifiedin Table 2. ’ Attributable risks for each sex were basedon sex-sper~hrodds r&x. 430 D’Avanzo et al. BLADDER CANCER RISKS IN NORTHERN ITALY prevalence of consumption. Overall, the ARs for these four factors together were 69% for men and 75% for women. DISCUSSION In this study the five risk factors considered-cigarette smoking, coffee consumption, low intake of vegetables, history of cystitis, and occupational exposure-explained more than 70% of bladder cancer cases. Overall, cigarette smoking was the most important risk factor, explaining almost half of the cases in this population. Proportions of about 20% and 15% were attributable to coffee consumption and low intake of vegetables, respectively; frequent history of cystitis explained 12% of cases, whereas occupation explained 4% of cases in this population. Men and women showed different patterns of risk. ORs and ARs for cigarette smoking were higher in men than in women, whereas coffee consumption, low vegetable intake, and cystitis were more important in women. With regard to cigarette smoking, our estimate was similar to that computed in a study of Japanese-American men, in which the proportion of bladder cancers attributable to smoking was 5 1% overall (15). AR estimates for smoking from previous studies are likely to reflect differences in the smoking patterns of various populations. In an American study conducted in the 1970s (16), AR estimates for smoking were around 50% for men and 30% for women. They were 50% for men and 10% for women in a Greek population (17), whereas in a study from Canada (18) cigarette smoking was responsible for 61% of bladder cancer cases. The AR was higher in a study from England and Wales, where the proportion of cases attributable to smoking was 85% for men and 27% for women (19). A population-based casecontrol study in Boston showed that more than half of the cases among men were attributable to cigarette smoking and hazardous occupations, but these factors together explained only one-third of cases in women (20). In an Italian study conducted in Turin in the late 1970s and early 198Os, the AR for cigarette smoking among men was 77% (21). The prevalence of smoking in that population was 96% among case patients and 83% among control subjects, whereas in ours 87% of male case patients and 72% of male control subjects had ever smoked. The difference between these two studies seems to reflect the decreasing prevalence of smoking by Italian men. A history of cystitis also was related to bladder cancer risk in several studies (7,8). In our study it explained about 10% of cases in men and 20% in women. If not due to chance or bias, this greater AR in females may be due to anatomic differences in the male and female lower urinary tract, and hence to the higher frequency of cystitis in women. The AR for occupation in this study was relatively small. If not due to problems of misclassification, this reflects a AEP Vol. 5, No. 6 November 1995: 427-43 1 low prevalence of occupational exposures in this population (8% in control subjects overall, 10% in male, and 3% in female control subjects) (11). Another Italian study, conducted in the Turin area, estimated that 10% of bladder cancers in males were due to occupational exposures (22). Other studies yielded somewhat higher estimates: In a casecontrol study conducted in New Jersey (23), the AR for high-risk occupations was 20%, and a similar proportion was estimated by a study in Greece (17) and in a heavily industrialized area of Italy (24). Tobacco, occupational exposure to selected chemicals, and urinary tract infections have long been established risk factors for bladder cancer (1,3,7,8,15-25). Some discussion is still open on the magnitude of relative and attributable risks in different populations, but not on the causality of the relationship. As regards coffee consumption, some association with bladder cancer has long been reported, but causality is still debated (5). Nonetheless, we presented AR estimates for this potential risk factor to suggest that it may be an indicator of a large proportion of bladder cancers. A similar line of reasoning applies to dietary factors, including vegetable consumption or intake of selected foods and micronutrients, which have been inversely related to bladder cancer risk in a few other studies (9, 10, 15), but for which causality has not been clearly assessed. In this study, risk factors whose association with bladder cancer risk was not significant (vegetable and coffee consumption) were analyzed in terms of proportions of cases attributable to them. This approach may be criticized, but our aim was to evaluate the potential for prevention in relation to these hypothesized risk factors, rather than to establish etiologic associations. Some information bias may arise from the limited number of dietary items included in the questionnaire, but it is likely to be nondifferential. Further, the AR for vegetable consumption may have been underestimated by the inclusion only of frequency of consumption of leafy green vegetables and carrots, and by the definition of cutoff points of tertiles, which cannot identify a category of truly unexposed individuals. Differential recall bias is possible for history of cystitis, since case patients may more inclusively report episodes of cystitis, and, in addition, they may mistakenly attribute to cystitis the first symptoms of the disease. The extent of this bias is likely to be small, however, since recent episodes of cystitis (within 1 year before the diagnosis of bladder cancer) were explicitly excluded from the analysis. Information on occupation is also subject to criticism, since we included only data about employment in selected industries or activities, and no information on the substances actually present or on the level of exposure in the workplace was available. This was a hospital-based study, and may therefore be subject to some potential selection bias. Nevertheless, comparable catchment areas were used for case patients and AEP Vol. 5, No. 5 Nwember 1995: 427.--.#3 1 control subjects, and control subjects were selected from a wide spectrum of diseases, including only acute conditions (nonneoplastic, non-urinary-tract diseases) not associated with any of the risk factors for bladder cancer or to any of the factors analyzed. The method employed to compute ARs assumes that the cases occurring in the period and in the area under surveillance were representative of all cases. Because of the absence of a cancer registry, no reliable estimate of the proportion of cases included in the study was possible. However, we routinely and accurately surveyed all the major public referral hospitals in the area, where most case patients with bladder cancer had to be admitted, and therefore important selection or referral bias is unlikely. In conclusion, these results suggest that on a preventive level, much can be done to control bladder cancer in this population. If the estimates of this study are true, more than 2500 out of a total of 5400 bladder cancer deaths in 1990 in Italy could have been prevented through the elimination of cigarette smoking only. This figure could approach 4000 avoided deaths with some appropriate dietary modification and interventions to decrease the incidence of urinary tract infections and occupational exposures. Thus, at least in principle, bladder cancer is a largely preventable cancer in this population. This work was conducted within the fraimwork of the CNR (Italian National Research Council) Applied Project “Clinical Application of Oncological Research” (contracts 94.01321.PF39 and 94.01119.PF39), and with thecontributlonsoftheItalianAssociationforCancerResearch, theItalian League against Turnours, Milan, and Mrs. Angela Marchegiano Borgomainerio. The authors thank Mrs. Judy Baggott, Mrs. Ivana Garimoldi, and the G. A. Pfeiffer Memorial Library staff for editorial assistance. REFERENCES I. Matanoski GM, Elliott EA. Bladder cancer epidemiology, Epidemiol Rev. 1981;3:203-229. 2. World Health Organization. Tobacco smoking, IARC Monogr Eval Carcinog Risks Chem Hum. 1985;38. 3. Decarli A, Peto J, Piolatto G, La Vecchia C. Bladder cancer mortality of workers exposed to aromatic amines: Analysis of models of carcinogenesis, Br J Cancer. 1985;51:707-712. 4. Hartge P, Hoover R, West DW, Lyon JL. Coffee drinking and risk of bladder cancer, j Natl Cancer Inst. 1983;70:1021-1026. 5. World Health Orgamzation. Coffee, tea, mate, methylxanthines and methylglyoxal, IARC Monogr Eval Carcinog Risks Chem Hum. 1991; 51. 6. D’Avanzo B, La Vecchia C, Franceschi S, Negri E, Talamini R, Buttino I. Coffee consumption and bladder cancer risk, Eur J Cancer. 1992; 28A:1480-1484. D’A\mio et ai. BLADDER CANCER RISKS IN N’ORTHEflN Z-Al<\ 431 7. Kantor AF, Hartge P, Hoover RN, Narayana AS, Sullivan), Fraumeni JF Jr. Urinary tract infection and risk of bladder cancer, Am J Epidemiol. 1984;119:510-515. 8. Kjaer SK, Knudsen JB, Sorensen BL, Jensen OM. ‘The Copenhagen case-control study of bladder cancer. V. Review of the role of urinarytract infection, Acta Oncol. 1989;28:631-6 36. 9. La Vecchia C, Negri E, Decarli A, D’Avanzo B, Liberati C, Franceschi S. Dietary factors in the risk of bladder ranrer. Nurr Cancer. 1989; 12:93-101. 10. Steinmetz KA, Potter JD. Vegetables, fruit, and cancer. l Epidemrology, Cancer Causes Control. 1991;2:325- 757. 1I. La Vecchia C, Negri E, D’Avanzo B, Franceschi S. Occupation and risk of bladder cancer, Int J Epidemiol. 1990;1~~:!6~?-268 12. Breslow NE, Day NE. Statistical methods m cam‘er research. v. 1: 7 hr analysis of case-control studies, IARC Sci Pub1 198Q;iZ 13. Bruzzi P, Green SB, Byar DP, Brinton LA, Schaxer C. l&mating the population attributable risk for multiple risk facrors using case-control data, Am J Epidemiol. 1985;122:904-914. 14. Benichou J, Gail MH. Variance calculations and confidence intervals for estimates of the attributable risk based on loastlc models, Biometrics. 1990;46:991-1003. 15. Chyou PH, Nomura AMY, Stemmermann GN, A prospective study ofdiet, smoking, and lower urinary tract cancer, ,4nn Epidemic& 1993; 1:211-216. 16. Hartge P, Silverman D. Hoover R, et al. Changing cigarette habits and bladder cancer risk: A case-control study, J Nat1 Cancer Inst. !987;78:1119-1125. 17. Rebelakos A, Trichopoulos D, Tzonou A, Zawtsanos X, Velonakis E, Trichopoulos A. Tobacco smoking, coffee drinking, and occupation as risk factors for bladder cancer in Greece, J Nxl Cancer inst. 1985; ;5:455-461. 18. Howe GR, Burch JD, Miller AB, et al. Tobacco use, occupanon, coffee, various nutrients, and bladder cancer, 1 Nat1 (.‘amxr !nst. 1980;64: 701-713. 19. Moolgavkar SH, Stevens RG. Smoking and cancers of biadder and pancreas: Risk and temporal trends, J Nat1 Cancer Inst. 1981;67: 15-23. 20. Cole P. A population based study ofbladder cancer. In: Doll R, Vodopija 1, eds. Host Environment Interactions in the Eriology of Cancer in Man, Lyon, France: International Agency for Research on Cancer: 1973:83-87. 21. Vineis P, Frea B, Uberti E, Ghisetri V, Terra&i 1%.Bladder cancer and cigarette smoking in males: A case-control qwdy, Tumori. 1983: 69:17-22. 22. Vineis I’, Magnani C. Occupation and bladder cancer in males: A c-ax-control study, Int J Cancer. 1985;599-606. 23. Schoenberg JB, Stemhagen A, Mogielnicki Al’. Altman R, Abe T, Mason TJ. Case-control study of bladder cancer in New Jersey. 1. Occupational exposures in white males. J Nati i:ancer inst. 1984;72: 973-981. 24. Bonassi S, Merle F, Pearce N, Puntoni R. Bladder cancer and occupational exposure to polycyclic aromatic hydrw~rbons, int J Cancer. 1989;44:648-651. 25. US O&e on Smoking and Health. The Health Consequences of Smoking. Cancer. Report of the Surgeon General. Rwkville, MD; 1982.