A PROSPECTIVE STUDY OF SMOKING AND RISK OF PROSTATE CANCER

Int. J. Cancer: 67,764-768 (1996) 0 1996 Wiley-Liss, Inc. Publicationof the InternationalUnion Against Cancer Publicationde I Union InternationaleCon...
Author: Guest
4 downloads 0 Views 581KB Size
Int. J. Cancer: 67,764-768 (1996) 0 1996 Wiley-Liss, Inc.

Publicationof the InternationalUnion Against Cancer Publicationde I Union InternationaleContre le Cancer

A PROSPECTIVE STUDY OF SMOKING AND RISK OF PROSTATE CANCER Hans-Olov ADAMri%3s7, Reinhold BERGSTROM~J, Goran ENCHOLMj, Olof NYREN',Alicja WOLK', Anders EKBOM'.~, Anders E N G L U N and D ~ John BARON'.^ Departments of 'Cancer Epidemiology and 2Statistics, Uppsala University, Uppsala, Sweden; 'Department of Epidemiology, Hatvard School of Public Health, Boston, MA, USA: 4NatiorialBoard of Health and Welfare, Stockholm; sNational Board of Occupational Safety and Health, Stockholm, Sweden; hDepartmerrtsof Medicine and Community and Family Medicine, Dartmouth Medical School, Hanover, NH, USA. We evaluated the hypothesis that smoking increases the incidence of and mortality from prostate cancer. High-quality smoking information was collected in 1971-1975 in a nationwide cohort of 135,006 male construction workers in Sweden. We achieved virtually complete follow-up through record linkages and ascertained as of December I99 I 2,368 incident cases of prostate cancer and 709 deaths due to this disease. Rate ratios (RR) of prostate cancer incidence and mortality, with 95% confidence intervals (CI), were estimated in Poisson-based age-adjusted models, with amount and duration of smoking as independent variables. We found no convincing association between current smoking status, number of cigarettes smoked or years since onset and risk of prostatic cancer. The ageadjusted incidence RR among previous smokers was I .09 and among current smokers I. I I compared with non-smokers. Weak and inconsistent trends were seen with increasing nurnber of cigarettes smoked per day and increasing duration among current smokers. Smokers of 15 or more cigarettes daily for at least 30 years experienced an incidence RR of 1.30. Mortality in ex-smokers was similar to that in never-smokers: it was, however, slightly increased among current smokers without any trend with amount smoked or duration. The weak and inconsistent associations between smoking and prostate cancer could easily have arisen due to bias or confounding. We therefore conclude that smoking is most likely not causally linked to the occurrence of prostate cancer.

el 1996 Wiley-Liss,Inc. Prostate cancer is the most common malignancy among men in many Western countries, but its causes are largely unknown (Tomatis, 1990; Nomura and Kolonel, 1991). This cancer is not generally considered to be tobacco-related (IARC, 1986), but available epidemiologic data are not entirely negative. While earlier studies of smoking generally found no association (Tomatis, 1990; Nomura and Kolonel, 1991; IARC, 1986; Hammond, 1966: Wynder et al., 1971; Armenian et al., 1975; Schuman et al., 1977), some (Honda et al., 1988; Hsing et aL, 1990, 1991; Hiatt et al., 1994), but not all (Ross et al., 1987; Hayeset al., 1994; Doll et al., 1994; van der Gulden et al.. 1994; Siemiatycki et al., 1995), recent investigations have indicated a modestly higher risk in smokers than in non-smokers. In particular, 2 large prospective studies reported an association (Hsing et al., 1990, 1991), whereas 1 did not (Doll et a/.. 1994). Use of smokeless tobacco, however, has consistently been associated with a slight excess in risk (Hsing et al., 1990, 1991; Hayes et al., 1994). A link between smoking and prostate cancer is biologically plausible. N-Nitroso compounds (IARC, 1986; Pour, 1983) or other components of tobacco smoke could be carcinogenic to the prostate, as for other internal organs such as the pancreas and the kidney (Silverman et al., 1994; McLaughlin et nl., 1995). A positive association could also have a hormonal basis since circulating levels of testosterone may be higher in smoking than in non-smoking men (Dai et a/., 1981, 1988; Deslypere and Vermeulen, 1984). These endocrine effects might also influence prognosis adversely, implying higher case fatality rate among smokers even in the absence of an etiologic role (Daniell, 1995). Because the adverse effects of cigarette smoking are widely known, retrospective assessment of smoking history in case-

control studies may be prone to bias, leading to either attenuation or exaggeration of true effects. Such bias is avoided in prospective studies, in which exposure is assessed prior to disease onset. We took advantage of a large prospective study in Sweden with detailed characterization of tobacco use. Various aspects of exposure could be studied in detail because smoking habits varied widely among cohort members, and we documented 2,368 incident cases of prostate cancer, with 709 deaths from this disease during follow-up. SUBJECTS AND METHODS

The cohort The Construction Industry's Organization for Working Environment Safety and Health (Bygghalsan), established in 1968, provided outpatient medical services to construction workers all over Sweden from 1969 through 1992 (Ostlund and Englund, 1971; Engholm ef al., 1987). The organization was a joint venture launched by construction trade unions and the Swedish Construction Employers' Association. The basic units were stationary and mobile clinics, typically staffed by a few nurses and a physician. The main activity was preventive health check-ups, offered to all blue- and white-collar workers in the building industry through regular (every 2nd year during the first years, every 3rd year thereafter) personal invitations and through visits or advertisements at virtually all major building sites. Beginning with visits in 1971, data from these health check-ups were compiled in a computerized central register. We considered 135,049 men who were first registered in 1971-1975. Forty-three individuals were excluded because they had prostatic cancer before entry to the cohort or due to inconsistencies in their data (death date or date of emigration before entry into the cohort), leaving 135,006 subjects for the final analyses. Their distribution by age and smoking status at start of follow-up and their accumulated person-years of cancer incidence follow-up are shown in Table I. Exposure information On average, each cohort member underwent 3 health check-ups. In the present analysis, however, we utilized only information obtained at the index visit (1971-1975), which defined entry to the cohort. Before each visit, the worker filled out a questionnaire with approximately 200 items, including a detailed smoking history. A t the clinic, the answers were double-checked by a nurse to minimize misunderstandings and inconsistent answers. Information on marital status and measured height and weight werc included in each record. The quality of smoking data has been reviewed previously (NyrCn et al., 1996). Briefly, missing data on smoking duration were noted in 894 current (1.3%) and 379 previous (1.4%) smokers. Perfect concordance between reports of smoking status 2-3 'To whom correspondence and reprint requests should be sent, at Department of Cancer Epidemiology, University Hospital, 751 85 Uppsala, Sweden. Fax: 46 18503431.

Received: March 26. 1996 and in revised form May 18, 1996.

765

SMOKING AND PROSTATIC CANCER RISK

Age at

Number of

Person-years

entry (yr) _ _ _ _ _

individuals ~

follow-up

Never

Ex

Current

450.906 440:436 167,312 2,369,006

7,386 27.6 31570 (31.1) 40,318 (29.9)

6;54i '24.4j 2,929 t25.6) 27,835 (20.6)

121793 '47.9j 4,953 i43.3) 66,853 (49.5)

< 30 30-39 40-49 50-59 > 60 All ages

_ _

~~

40,559 39,911 25.364 26:720 11,452 135,006

Number (7o)by smoking status at ently

~~

years apart was found in 89% of cases. Inconsistencies (e.g., subjects who indicated that they were current or former smokers in the first questionnaire and who asserted that they had never smoked in the second questionnaire) were found in 2.6% of cases. Follow-lip Major efforts were made to ensure that the national registration numbers, unique personal identifiers assigned to all Swedish residents, were complete and valid. The proportion with incorrect national registration numbers was less than 1 in 1,000. The national registration number was used for follow-up through record-linkage to the National Death Registry, Migration Registry and Cancer Registry. The Cancer Registry, founded in 1958, has coded malignant neoplasms according to the International Classification of Diseases, 7th edition (ICD-7). during the entire period of study. Registration of newly diagnosed cancers is considered to be at least 96% complete, and about 98% of all prostate cancers were confirmed morphologically (by histopathology or cytology). Information on stage of disease is not included in the Registry (Cancer Registry, 1992). The incidence of prostatic cancer (ICD-7:177) and mortality with this malignancy coded in the Death Registry as the underlying cause of death, are reported here. Each cohort member contributed person-years from the date of first registered visit until the date of a prostatic cancer diagnosis (in the incidence analyses), death, emigration or December 31, 1991, whichever occurred first. The number of person-years accumulated in each stratum defined by age at entry is shown in Table I. The average age at entry to the cohort varied between 38.9 and 41.8 years in the 3 categories of never-, previous and current smokers. Altogether 2,369,006 person-years were observed in the analyses of prostatic cancer incidence and 2,377,960 in the corresponding analyses of mortality.

Statistical analyses Because prostatic cancer risk is highly age-dependent, we first analyzed the age distribution in different exposure categories. Then, the numbers of cases and person-years experienced were computed within 12 age intervals (80 years) for each category of the exposure variables studied. These data were used to obtain age-adjusted rate ratios (RRs) in a first simple multiplicative multivariate model, with age and relevant exposure as independent variables. Due to the small risk of prostatic cancer at younger ages, age was included with 9 catcgories ( < 45.45-49 . . .75-79, > 80) in the actual modeling. Duration of smoking in current smokers was counted only until the date of entry since we had no access to information on those who quit after this date. Modeling was based on the assumption that the number of cases followed a Poisson distribution. Models were fit using the maximum likelihood method (McCullagh and Nelder, 1989). On the basis of the results thus obtained, further multivariate modeling was then performed, adjusting for variables such as body mass index and marital status. We also estimated further

multivariate models where age at entry to the cohort was included as an explanatory variable, in addition to the basic variables actual age during follow-up and exposure status. The purpose of this modeling was to adjust for possible birth cohort effects. As recruitment to the construction workers cohort occurred during a limited time (1971-1975), age at entry to the cohort is closely related to birth cohort.

RESULTS

We identified a total of 2,368 men in the cohort who developed prostatic cancer during follow-up through 1991. The age-adjusted RRs for smoking-related variables are given in Table 11. These were obtained from multivariate Poisson models, with age included in addition to each of the variables shown. The fact that prostatic cancer risk is strongly agedependent and that several of the explanatory variables were related to age means that the age-adjusted relative risks shown in many cases deviate substantially from the unadjusted relative risks that can be inferred from the number of cases and person-years shown in Table 11. There were slightly elevated risks for previous and current smokers compared with neversmokers, the point estimates being 1.09 and 1.11, respectively. We found a weak relationship between number of cigarettes smoked per day and risk of prostatic cancer ( p for trend = 0.04). Among smokers of more than 14 cigarettes, R R was 1.12 (95% CI 0.98-1.27); however, in smokers of 25 or more cigarettes per day, the risk was unity. Table I1 also shows relative risk estimates by duration of smoking at entry to the cohort among ex-smokers and current smokers. Relative risks among smokers were generally slightly larger than 1, but in only 2 categories was the risk increase statistically significant at the 5% level. We found no obvious trend with duration among ex-smokers or current smokers, though a formal linear trend test yielded a significant value for the latter @ = 0.05). There was no trend seen among pipe smokers, though an increased risk was seen in men smoking more than 100 g pipe tobacco per week ( R R = 1.53;Table 11). Snuff dipping was unrelated to prostatic cancer (Table II), while cigar smoking was too infrequent for meaningful analysis. We considered separately the possible effects of smoking among men who had been smokers for 30 years or longer at entry into the cohort. Compared with never-smokers, RRs were 1.04 and 1.06 among smokers of 1-4 and 5-14 cigarettes per day, respectively. Men who smoked 15 or more cigarettes daily had an R R of 1.30 (95% CI 1.06-1.59). Body mass index (BMI) and marital status had weak associations with risk of prostatic cancer. Addition of these variables to the statistical models yielded adjusted RRs closely similar to those shown in Table 11. In our study design, duration of smoking at entry to the cohort is correlated with age at entry and thereby with birth cohort. We analyzed whether birth cohort effects on prostatic cancer incidence unrelated to smoking might have confounded the analyses of RRs by duration. Birth cohort effects were, however, minimal

766

ADAMI ETAL.

compared with the age effect, and inclusion of these effects did not appreciably change the effects associated with smoking. Former smokers had essentially the same risk of dying from prostatic cancer as never-smokers ( R R 1.03), while current smokers had a statistically significant 26% excess risk (Table 111). However, we found no trend with increasing number of cigarettes smoked per day, nor any association with pipe smoking or snuff dipping. Ex-smokers with varying durations of smoking had risks closely similar to that in never-smokers. Among current smokers, we found a general 1 9 4 1 % excess mortality rate from prostatic cancer but without any trend with duration of smoking (Table 111). The formally significant trend effect is a reflection of the slightly higher risk among all current smokers, a risk increase that is not associated with duration. DISCUSSION

In this large prospective study, we found a weak, but statistically significant, positive association between smoking and the risk of developing prostatic cancer. Among smokers, the mortality R R was slightly higher than the incidence RR, but we found no trend with amount smoked or duration of smoking. Although early studies of smoking and prostatic cancer largely showed no association (Tomatis, 1990; Nomura and

Kolonel, 1991; IARC, 1986; Hammond 1966; Wynder et al., 1971; Armenian et al., 1975; Schuman et al., 1977), some of these investigations had potential shortcomings that may have led to a conservative bias Some of the studies were small and had limited statistical power. In the case-control studies, a weak or moderate association might have been overlooked because of the difficulties associated with random misclassification of exposure, non-response or the use of hospital controls. Cohort studies of prostatic cancer mortality could not distinguish an effect of smoking on incidence from an effect on prognosis because both would influence mortality. In contrast to these early investigations, several recent case-control (Honda et al., 1988; Hayes et al., 1994) and cohort (Hsing et al., 1990, 1991) studies have suggested a weak positive association between cigarette smoking and prostatic cancer incidence (Honda et al., 1988; Hayes et al., 1994) or mortality (Hsing et al., 1990, 1991). Strengths of our study include careful data collection with a combination of self-administered questionnaires and face-toface interviews by nurses. The consistent responses in repeated interviews attest to the high quality of the smoking data. Moreover, repeat measurements indicated that only few men who were non-smokers at entry later took u p smoking and that only few smokers, notably few healy and long-term smokers, quit smoking during follow-up. Although the reporting of

TABLE I1 - AGE-ADJUSTED RATE RATIO ( R R ) WITH 95% CONFIDENCE INTERVALS (CI) OF PROSTATE CANCER INCIDENCE THROUGH I Y Y I AMONG SWEDISH CONSTRUCTION WORKERS IN RELATION TO SMOKING HISTORY AT ENTRY TO THE COHORT, 1971-1975 Variable

Smoking status] Never Ex Current Number of cigarettesiday 0 1 4 5-14 15-24 > 25

Duration of smoking among ex-smokers (Yr) Never smokers 1-10 11-20 > 21

Duration of smoking among current smokers (yr) Never smokers 1-10 11-20 21-30 31-40 > 41

Pipe, g tobaccoiweek Never?

< 30

30-100 > 100

Snuff dipping Never1 Ever

Personyears

Number

RR

of cases

(95?4,CI)

40,318 27,835 66,853

717,072 489,178 1,162,756

682 617 1,069

1.00 (ref) 1.09 0.98-1.22 1.11l1.01-1.231

63,188 13,552 34,007 21,653 2,565

1,103,480 236,905 602,063 381,276 44,239

1,348 282 459 239 38

1.00 (ref)

40,318 12,405 8,789 6,163

717,072 225,809 155,263 101,498

682 143 199 273

1.00 ref) 1.20 11.00-1.44) 1.14 0.97-1.34 1.03 [0.90-1.19) p for trend: 0.33

40,3 18 17,885 20,527 14,145 10,168 3,234

717,072 327,840 370,880 243,221 160,260 45,482

682 19 92 255 454 228

1.00 (ref) 0.68 (0.43-1.07) 1.27 1.02-1.58 1.09 0.94-1.26

100,078 14,626 19,580 696

1,767,159 255,321 334,739 11,370

1,543 320 476 29

103.218 31;788

1.808.836 560,170

1,840 528

Number of subjects

p for trend: 0.04

$

1

p for trend: 0.05 1.00 (ref) 1.08 (0.96-1.22) 0.98 0 8&108 1.53 [1:06-2:21)) p for trend: 0.62

~

'Includes cigarette, pipe and cigar smoking.-?Includes cigarette and cigar smoking.

767

SMOKING AND PROSTATIC CANCER RISK

TABLE Ill - AGE-ADJUSTED RATE RATIO (RR) WITH 95% CONFIDENCE INTERVALS (Cl) OF PROSTATE CANCER MORTALITY THROUGH 1991 AMONG SWEDISH CONSTRUCTION WORKERS IN RELATION TO SMOKING AT ENTRY TO THE COHORT, 1971-1975

Smoking status Never Ex Current Number of cigiday 0 1-4 5-14 15-24 > 25

198 168 343

1.00 ref) 1.03 tO.84-I 3 3 1.26 (1.06-1.50{

411 80 141

1.00 ref) 0.99 [0.78-1.26) 1.13 (0.93-1.37) 1.05 (0.82-1.35)

';]

p for trend: 0.34

Duration of smoking among ex smokers (yr) Never smokers 1-10 11-20 > 21 Duration of smoking among current smokers (yr) Never smokers 1-10 11-20 21-30 3 1-40 > 41

198 39 52 76

1.00 (ref) 1.12 (0.80-1.58) 1.06 (0.78-1.44) 0.98 (0.75-1.28) p for trend: 0.96

198

1.00 (ref) 1.41 (0.99-1.99)

if]

p for trend: 0.03

Pipe, giweek Never < 30 30- 100 > 100

456 106

1.00 (ref) 1.21 (0.98-1.50) 1.00 (0.83-1.21) p for trend: 0.26

Snuff dipping Never Ever

553 156

1.00 (ref) 0.99 (0.8i-1.18)

';f)

'Numbers of subjects and person-years are closely similar to those shown in Table 11. cigarette smoking was generally quite good, errors of recall or uptake of smoking after the questionnaire was completed could have led to some underestimation of the true association. However, any trend with duration of smoking (or latency since start of smoking) would be over-estimated since exposure information was taken into account only until the date of entry into the cohort, while many current smokers continued their smoking habits throughout the follow-up period. Bias due to differential misclassification of exposure is inconceivable with our study design. The zealous checks of the national registration numbers among cohort members and the almost 100% completeness of the Swedish Cancer Register (Cancer Registry, 1992) ensured virtually complete follow-up for an average of about 18 years. However, since smoking causes a myriad of health problems (Doll rt al., 1994), many of which require recurrent (eg., peptic ulcer) or long-term (e.g., cardiovascular and lung diseases) treatment, smokers may have been under closer medical surveillance than non-smokers. In Sweden, digital rectal examination of the prostate is routinely performed as part of a physical examination in men. Because of the prevalence of early stage, asymptomatic prostatic cancer is relatively high in aging men (Gustafsson et al., 1992), more frequent palpation of the prostate in smokers than in non-smokers may generate a spuriously inflated incidence in smokers. A counteracting bias

would occur if non-smokers are more health-conscious than smokers and therefore frequently undergo routine rectal digital examination. Such surveillance bias may be due both to over-diagnosis of lesions that would otherwise never have been diagnosed and to earlier diagnosis of clinically overt cancers. No treatment of prostatic cancer has a documented beneficial effect on survival (Whitmore, 1994); consequently, it is not at all clear that earlier diagnosis would reduce mortality through improved prognosis. In an observational study as large as ours, the effect of co-variates could easily overwhelm random variation (Taubes, 1995). Some of the weak associations attained formal statistical significance, but this implies little about causality and should not be over-interpreted; these associations may be due to confounding or bias. Control of confounding is difficult in studies of prostatic cancer because the causes of this malignancy are largely unknown. It is easy, however, to imagine that smokers differ from non-smokers in many ways, ranging from physical exercise to dietary habits, alcohol consumption and drug use (Subar and Harlan, 1993). Confounding by these and other factors that we could not account for might easily have produced the excess risks in smokers even in the absence of a causal association. The more important issue is whether confounding can lead to an underestimation of the true excess risk because exposure to some protective factor(s) was more common in smokers than in non-smokers. Among the few factors suspected (Nomura and Kolonel, 1991) to be associated with prostatic cancer, none could plausibly negatively confound the association with cigarette smoking to such a degree that a substantial excess risk is almost cancelled. Among current smokers, the excess risk for mortality was slightly, but consistently, higher than that for incidence. Although this difference may be due to chance, it is notable that previous prospective studies reporting a positive association between smoking and prostatic cancer analyzed mortality, not incidence (Hsing et al., 1990, 1991). The cause of death is not always obvious in patients with prostatic cancer. notably not in subjects who, like smokers, have other serious health problems. For example, an inflated mortality RR would occur if some deaths in smokers, that are in reality due to cardiovascular or other smoking-related diseases, were classified as being due to prostatic cancer. Moreover, the adverse, mainly thromboembolic, effects of hormonal treatment with estrogens (Johansson et af., 1991) might occur more frequently in smokers than in non-smokers. When such complications are fatal, prostatic cancer may be classified as an underlying or contributory cause of death. Hence, bias due to differential misclassification of death between smokers and non-smokers with prostatic cancer might generate a spuriously inflated mortality RR. In summary, we are faced with a small excess risk of borderline statistical significance and without convincing doserisk trends. Potential confounding and bias seem more likely to have entailed slightly exaggerated, rather than under-estimated, relative risks in smokers. Several important criteria for casuality, notably strengths of association and dose-response, were lacking in our data. W e conclude, therefore, that our study provides evidence that smoking is not substantially related to the occurrence of prostatic cancer.

ACKNOWLEDGEMENTS

The authors acknowledge the decisive support of Prof. Nils Stjernberg during the initiation of this study. Dr. Stjernberg's sudden and untimely death in June 1995, a profound shock to us, prevented his participation in the completion of this work. This work was supported by grants from the Swedish Cancer Society.

768

ADAMI ETAL.

REFERENCES ARMENIAN, H.K., LILIENFELD, A.M., DIAMOND, E.L. and BROSS, JOHANSSON, J.E., ANDERSSON, S.O., HOLMBERG, L. and BERGSTROM, I.D.J., Epidemiologic characteristics of patients with prostatic neo- R.. Primary orchidectomy versus estrogen therapy in advanced prosplasms. Amer. J. Epidemiol.. 102,47-54 (1975). tatic cancer-a randomized study: results after 7 to 10 years of CANCER REGISTRY, Cancer incidence in Sweden 1989, National Board follow-up.J. Urol., 145,510-523 (1991). of Health and Welfare, Stockholm (1992). MCCULLAGH, P. and NELDER.J.A., Geiieralized linear models, Chapman and Hall, London (1989). DAI, W.S., GUTAI, J.P., KULLER, L.H. and CAULEY, J.A., Cigarette smoking and serum sex hormones in men. Amer. J. Epidemiol., 128, MCLAUGHLIN, J.K., LINDBLAD, P., MELLEMGAARD, A.: MCCREDIE, M., 796-805 (1988). MANDEL,J.S., SCHLEHOFER, B., POMMER,W. and ADAMI,H.O., DAI,W.S., KULLER, L.H., LAPORTE, R.E., GUTAI,J.P., FALVOGERARD,International renal-cell cancer study. 1. Tobacco use. bit. J. Cancer, 60, L. and GAGGIULA, A,, The epidemiology of plasma testosterone levels 194-198 (1995). in middle-aged men.Amer. J. Epidemiol., 114,804-816 (1981). NOMURA,A.M.Y. and KOLONEL,L.N., Prostate cancer: a current DANIELL, H.W., A worse prognosis for smokers with prostate cancer.J. perspective. Epidemiol. Rev.. 13,200-227 (1991). Urol., 154,153-157 (1995). R., NYSTROM, L., ENGHOLM, G., EKBOM, A,, NYREN,O., BERGSTROM, DESLYPERE, J.P. and VERMEULEN, A., Leydig cell function in normal ADAMI, H.O. and STJERNBERG, N., Smoking and colorectal cancer-a men: effect on age, life-style, residence, diet and activity. J. clirt. 20 year follow-up study of Swedish construction w0rkers.J. nut. Cancer Endocrinol. Mefab., 59,955-962 (1984). Inst. (1996) (In press). DOLL,R., PETO,R., WHEATLY, K., GRAY,R. and SLTHERLAND, J., OSTLUND, E. and ENGLUND, A., Occupational safety and health in the Mortality in relation to smoking: 40 years observations on male British Swedish construction industry. Work Environment Health, 8, 27-32 doctors. Brit. med. J., 309,901-911 (1994). (1971). ENGHOLM, G., ENGLUND, A.. FLETCHER,A.C. and HALLIN,N., Respiratory cancer incidence in Swedish construction workers ex- POUR,P., Prostatic cancer induced in MRC rats by N-nitroso bis posed to man-made mineral fibres and asbestos. Ann. occiip. Hyg.. 31, (2-oxopropy1)amine and N-nitroso bis(2-hydroxyproby1)amine. Carcinogenesis, 4,49-55 (1983). 663-675 (1987). H., PAGANINI-HILL, A., HONDA. G. and HENDERGUSTAFSSON, O., NORMING, U., ALMGARD, L.E., FREDRIKSSON, A,, Ross, R.K., SHIMIZU, GUSTAVSSON, G., HARVIG, B. and NYMAN, C.R., Diagnostic methods SON, B.E., Case-control studies of prostate cancer in blacks and whites in southern California. J. nut. Cancer Insf., 78,869-874 (1987). in the detection of prostate cancer: a study of a randomly selected population of 2,400 men. J. Urol., 148, 1827-1831 (1992). C., BAIJER, H., SCARLkTT. J. SCHUMAN, L.M., MANDEL, J., BLACKARD, HAMMOND, E.C., Smoking in relation to the death rates of one million and MCHUGH,R., Epidemiologic study of prostatic cancer: preliminary report. Cancer Treat. Rep., 61,326-331 (1977). men and women. J. nut. Cancer Insf. Monogr., 19,127-204 (1966). HAYES,R.B., POTTERN, L.M., SWANSON, G.M., LIFF, J.M., SCHOEN- SIEMIATYCKI, J., KREWSKI, D., FRANCO, E.L. and KAISERMAN, M., BERG,J.B.. GREENBERG, R.S.. GROSSBART SCHWARTZ, A., MORRIS Associations between cigarette smoking and each of 21 types of BROWN.L., SILVERMAN, D.T. and HOOVER,R.N., Tobacco use and cancer; a multi-site case-control study. Inf. J. Epidemiol., 24, 504-514 prostate cancer in blacks and whites in the United States. Cancer (1995). Causes Control, 5,221-226 (1994). SILVERMAN, D.T. and 13 OTHERS,Cigarette smoking and pancreas HIATT, R.A., ARMSTRONG, M.A., KLATSKY, A.L. and SIDNEY,S., cancer: a case-control study based on direct interviews. J. nut. Cancer Alcohol consumption, smoking and other risk factors and prostate Inst., 86, 1510-1516 (1994). cancer in a large health plan cohort in California. Cuncer Causes SUBAR,A.F. and HARLAN, L.C., Nutrient and food group intake by Control, 566-72 (1994). tobacco use status: the 1987 national health inteiview survey. Ann. HONDA,G.D., BERNSTEIN, L., Ross, R.K.. GREENLAND, S., G ~ R K I N S N.Y. , Acad. Sci., 686,310-322 (1993). V. and HENDERSON, B.E., Vasectomy, cigarette smoking and age at first sexual intercourse as risk factors for rostate cancer in middle- TAUBES,G., Epidemiology faces its limits. Science, 269, 164-169 (1995). aged men. Brit. J. Cancer, 57,326-331 (19885. HSING,A.W., MCLAUGHLIN, J.K., HRUBEC,Z., BLOT, W.J. and TOMATIS,L. (ed.), Cancer: causes, occurrence and control, IARC FRAUMENI, J.F., Tobacco use and prostate cancer: 26-year follow-up of Scientific Publication 100, IARC, Lyon (1990). US veterans.Amer. J. Epidemiol., 133,437-441 (1991). V A N DER GULDEN, J.W., VERBEEK, A.L. and KOLK,J.J., Smoking and HSING, A.W., MCLAUGHLIN, J.K., SCHUMAN, L.M., BJELKE, E., GRID- drinking habits in relation to prostate cancer. Brit. J. Urol., 73,382-389 S., CHIEN,H.T. and BLOT,W.J., Diet, tobacco LEY,G., WACHOLDER, (1994). use, and fatal prostate cancer: results from the Lutheran Brotherhood WHITMORE, J.F.. Localised prostatic cancer: management and deteccohort study. Cancer Res., 50,6836-6840 (1990). tion issues. Lancet, 343, 1263-1267 (1994). IARC, Monographs on the evahiatioti of carcinogenic risks of chemicals EL., MABUCHI. K. and WHITMORE, W.F., Epidemiology Of f o humans. Tobacco smoking, pp. 199-298, International Agency for WYNDER, cancer of the prostate. Cancer, 28,344-360 (1971). Research on Cancer, Lyon (1986).

Suggest Documents