Secondary Logo

Journal Logo

Joint Effects of Smoking and Body Mass Index on Prostate Cancer Risk

Sharpe, Colin R.1; Siemiatycki, Jack1,2


Between 1979 and 1985, a population-based case-control study of cancer at multiple sites was carried out in Montréal. A total of 399 cases with histologically confirmed prostate cancer and 476 population controls, 45–70 years of age, gave face-to-face interviews and provided adequate smoking histories. We analyzed the effects of smoking cigarettes only and of smoking cigars, or pipes, or both, with or without cigarettes, on the risk of prostate cancer. Overall, the associations between smoking cigarettes and prostate cancer were weak and compatible with no effect; the associations with cigar and pipe smoking were stronger. Among men with high body mass index, however, we found appreciable associations between cigarette smoking and prostate cancer risk. A history of ever smoking daily was associated with an odds ratio of 2.31 (95% confidence interval = 1.09–4.89). Risk increased with the amount smoked per day and with the duration of smoking. Taken together, the findings of increased risk associated with cigar and pipe smoking and the findings of increased risk associated with cigarette smoking among obese men suggest that tobacco smoking may be a risk factor for prostate cancer.

From 1Institut national de la recherche scientifique (INRS)-Institut Armand-Frappier, Université du Québec, Laval,

2Joint Departments of Epidemiology and Biostatistics and of Occupational Health, McGill University, Montréal, Québec, Canada.

Address correspondence and reprint requests to: Dr. Jack Siemiatycki, Epidemiology & Biostatistics Unit, Institut Armand-Frappier, 531, boulevard des Prairies, Laval, Québec, Canada H7V 1B7.

This project has been supported by the National Health Research and Development Program of Canada, the National Cancer Institute of Canada, the Institut de Recherche en Santé et Securité au Travail (Québec), and Health Canada. C. R. Sharpe is supported by a Postdoctoral Fellowship from the Canadian Institutes of Health Research. Jack Siemiatycki is supported by a Distinguished Scientist Award from the Canadian Institutes of Health Research.

Submitted March 22, 2000; final version accepted November 28, 2000.

Autopsy studies of latent prostate cancer have shown that the prevalence of noninfiltrative disease varies little by race or by country, whereas incidence rates of clinically evident prostate cancer vary considerably. 1 This pattern suggests that promotion or progression of prostate cancer is related to exogenous factors. 2

Evidence regarding the effects of smoking on prostate cancer incidence has been weak and inconsistent. A consensus conference on smoking and prostate cancer concluded that there was inadequate evidence to relate smoking causally to the occurrence of prostate cancer. 3 The report recommended, however, that future studies attempt to identify subgroups that may be susceptible to smoking.

Tobacco smoking may be a factor that promotes prostate cancer or causes it to progress. A recent study indicated that cigarette smoking during the 10-year period preceding diagnosis was associated with the development of both metastatic and fatal prostate cancer. Although smokers may present later in the evolution of their disease, smoking could lead to more aggressive disease. 4 Such effects have been observed with other cancers. 5

The present analysis aimed to illuminate the relations between smoking various types of tobacco and the development of prostate cancer, using data from a previously conducted case-control study. In addition to cigarette smoking data, we had data on subjects’ consumption of cigars and pipe tobacco. Furthermore, because the potency of carcinogenic agents tends to be sensitive to dietary factors, 5 we reasoned that the effects of tobacco smoking could be modified by body mass index (BMI). Increasing BMI, an indicator of adiposity, is weakly associated with an increasing risk of prostate cancer and more strongly with prostate cancer mortality. 6 These relations led us to examine whether the effects of smoking might be more evident among particular strata of BMI.

Back to Top | Article Outline

Subjects and Methods


The overall design of the study has been described in detail elsewhere. 7,8 Between 1979 and 1985, virtually all (97%) incident, histologically confirmed cancer cases at 21 sites occurring in men 35–70 years of age who were diagnosed at all of the large hospitals in metropolitan Montréal were accrued. Although the primary objective was to assess occupational exposures, the data collection included additional variables, some of which were used for the present analysis. Of 557 eligible, incident cases of prostate cancer (code 185, International Classification of Diseases, 9th revision), data were obtained from 449 (80.6%) either by face-to-face interviews of the patients or proxies or by a less extensive self-administered questionnaire given to those unwilling to be interviewed. This study was restricted to the 399 cases who underwent the face-to-face interviews and who provided adequate smoking histories (age range, 47–70 years). By examining information obtained from the medical records of all cases, Richardson 9 showed that there was virtually no nonresponse bias related to age, income, ethnicity, marital status, cigarette smoking, and alcohol consumption.

During the same period, 740 population controls were selected, in some years from electoral lists and in others by random digit-dialing, of whom 533 were interviewed (72.0%). We used the data from the 476 controls who underwent interviews and who were between 45 and 70 years of age on the assigned date of pseudodiagnosis, based on the year of interview.

Back to Top | Article Outline

Exposures and Covariates

Subjects were asked whether they smoked cigarettes, cigars, or a pipe nearly every day. If so, they were asked the age at which they began, the age at which they stopped (for ex-smokers), and the average daily amount smoked. We translated intensity and duration of cigarette smoking into a cumulative exposure variable, defined as the product of the average number of packages of cigarettes smoked per day (20 cigarettes per pack) and the duration of smoking in years and expressed as “pack-years.” Similar variables were constructed for cigar smoking using the number of cigars smoked per day (“cigar-years”) and for pipe smoking using the number of pipes smoked per day (“pipe-years”). Categories of these cumulative exposure variables were created by dividing the distributions among exposed controls into approximate tertiles.

We calculated BMI [weight (kg)/height squared (m2)] using the subjects’ usual weight when in good health. Subjects were also asked about their consumption of beer, wine, and spirits. For the subjects who drank daily, we constructed a composite exposure variable (“drink-years”) to express the cumulative exposure to alcohol by summing the cumulative exposure variables for each beverage. 10 In our analyses, we classified subjects into three categories: those who never drank weekly, those who drank weekly or daily with a cumulative exposure ≤30 drink-years, and those who drank daily with a cumulative exposure >30 drink-years. Because some of the cases were interviewed in the hospital and others at home and all of the controls were interviewed at home, it was impossible to blind the interviewers to the subjects’ disease status. We calculated odds ratios (ORs) for prostate cancer by aspects of cigarette smoking to estimate incidence density rate ratios, while correcting for potential confounders represented as categorical variables, with unconditional logistic regression, 11 fitted with the SAS LOGIST procedure. 12

Back to Top | Article Outline


Table 1 shows the distributions of the cases and controls by the levels of various attributes. The mean age of the cases was 63.0 years (standard deviation, 5.0); for the controls it was 60.3 years (standard deviation, 6.6). The subjects were mainly of French origin. Of those who were not French, most were of other European origins; only 3.3% of cases and 4.4% of controls were of non-European origin. Table 2 shows the corresponding distributions by BMI; the cutpoints were determined from the tertiles among the controls. The table also shows ORs for prostate cancer by BMI, with the middle category of BMI as the reference group. Risk tended to increase with increasing BMI, although the trend was not strong.

Table 1

Table 1

Table 2

Table 2

Table 3 shows the frequencies of the eight possible combinations of the three types of tobacco used. Each lifelong pattern of tobacco use was represented as a binary variable, indicating “ever” or “never” smoked. Cigarettes were by far the most frequently used type. The cigar and pipe smokers tended to have smoked cigarettes as well. All types of tobacco use were associated with ORs >1.0. The combinations that included cigar smoking were associated with the largest excess risks.

Table 3

Table 3

Table 4 shows ORs for prostate cancer by cumulative exposure to the smoking of cigarettes, cigars, and pipes. The results for each type of tobacco smoking are presented either without or with adjustment for the other two types. All three types of tobacco smoking showed evidence of weak associations with prostate cancer. The ORs were higher for cigar smoking than they were for cigarette or pipe smoking.

Table 4

Table 4

To examine the joint effects of tobacco smoking and BMI on prostate cancer risk, we chose to restrict our analysis to subjects who smoked only cigarettes to prevent the misclassification of exposure that could arise from either ignoring the use of the other types of tobacco or from attempting to construct composite tobacco exposure variables.

Table 5 shows ORs for prostate cancer associated with exclusive cigarette smoking, using as the reference group nonsmokers of any tobacco; that is, smokers of cigars and pipes were excluded from this and the following analyses. There was a relatively small increase in risk associated with the highest level of exposure, expressed in terms of pack-years. The 95% confidence intervals for all estimates were broad, however.

Table 5

Table 5

To examine whether the effects of cigarette smoking were modified by BMI, we repeated the analyses shown in Table 5 within levels of BMI. We found that the effects of smoking were similar in the two strata pertaining to the lower two tertiles of BMI, so we combined these strata.

Table 6 shows the results of this analysis. The ORs refer to a common reference group consisting of the men in the lower category of BMI who never smoked. Although in the lower stratum of BMI smoking was associated with ORs <1.0, the data were also reasonably compatible with no effect. Furthermore, there was no tendency for the ORs to increase with increasing pack-years of cigarette smoking. Nonsmokers in the higher stratum of BMI had a much lower risk than the reference group, whereas “ever smoking” was associated with an OR equal to that of the reference group. The ORs increased steadily, with increasing exposure expressed as pack-years.

Table 6

Table 6

Because the ORs in the higher stratum of BMI tended to increase monotonically with increasing duration and amount smoked per day (results not shown), we have presented the results in Table 6 using another reference group consisting of the men who never smoked with BMI >26.660. This approach allowed us to estimate the risk of cigarette smoking relative to not smoking among men of similar BMI. Table 7 shows that among the men in the highest tertile of BMI, cigarette smoking was associated with prostate cancer; the patterns by duration, amount, and quitting were those that would be expected of a causal relationship.

Table 7

Table 7

Back to Top | Article Outline


Consistent with the results of other studies, we found weak positive first-order associations between tobacco smoking and prostate cancer (Tables 3 and 4). 3,13 The associations appeared stronger for pipe and, especially, cigar smoking than for cigarette smoking. It was not possible to study the effects of cigar and pipe smoking further. It was possible, however, to examine further the risk associated with cigarette smoking, and we found a striking interaction between BMI and cigarette smoking (Table 6). Ostensibly, the risk of prostate cancer was lower among nonsmokers with high BMI than among nonsmokers with low BMI, the reference group. Among smokers with high BMI, the risk was equivalent to that of the reference group, whereas among smokers with low BMI the risk appeared to be less than that of the reference group.

One could attempt to interpret these results using a common reference group, or one could examine them by rows to study the modification of the effect of smoking by BMI, or by columns to study the effects of smoking among men with similar BMI. Adopting the latter perspective, among men with high BMI, smokers experienced an increased risk of prostate cancer (Table 7). This conclusion rests on the observation that risk was lower among nonsmokers with high BMI than among nonsmokers with low BMI, the reference group (Table 6). Although these results could have occurred by chance or could have been manifestations of bias, they might also reflect the underlying biological relations between BMI, smoking, and blood levels of sex hormones, which will be discussed below. Among men with low BMI, smoking appeared to be protective (OR = 0.70; 95% confidence interval = 0.41–1.20) (Table 6). Because the confidence interval was wide, this result is also compatible with smoking having no effect among men with low BMI. It could also have occurred by chance or could have been due to bias. Alternatively, it could reflect the underlying biological relations. Before considering possible biological explanations of our results, we will consider the limitations of the study and possible biases.

Although exposure to tobacco and to the covariates may have been misclassified to some extent, it is unlikely that misclassification of any of these variables could explain the effect modification by BMI that we observed. The quality of the data derived from proxies may have differed from that derived from the subjects themselves. We had few proxy respondents in this analysis, however, and we adjusted for respondent status. The methods of subject accrual ensured that the case and control series were population-based. Although selection bias could have occurred because of nonresponse, it did not appear to be appreciable. 9 We doubt that recall bias was appreciable, because the interviews elicited information about regular, repeated exposures to tobacco, which are usually not difficult to recall.

Control of confounding is problematic in studies of prostate cancer, because only a small fraction of cases can be attributed to known determinants. 14 Recognized determinants include age, race, family history, 14 and country of residence. 1,2 Confounding by race was unlikely, because almost all subjects were of European origin. We had no information about family history, but it is unlikely that there would have been a strong relation between the confounder (family history) and the exposure variable (smoking) in only one specific stratum of BMI.

Because a high-fat diet may be a determinant of prostate cancer 15–17 and because cigarette smoking tends to be associated with a high-fat diet, 18,19 it is possible that our results could be confounded by dietary fat. Given that smoking tends to increase smokers’ energy expenditure, 20 it is possible that smokers among the highest tertile of BMI may have increased their caloric intake by increasing their fat consumption to maintain their body weight to a greater extent than smokers among the lower tertiles.

Because BMI is inversely associated with energy expenditure from physical activity, it is also possible that the effect modification we observed could be related to levels of physical activity if it is a determinant of prostate cancer risk. The evidence for this relation is conflicting, 21–24 however, and thus, it is unclear how physical activity might be involved.

Back to Top | Article Outline

Interpretation of Results

Because tobacco smoke is known to have tumor-initiating, -promoting, and -progressing effects on the development of other cancers, 5 it is plausible that it could have similar effects on prostate cancer. The increases in risk observed with increasing duration and amount of cigarette smoking and the decrease observed after quitting ≥10 years before diagnosis among the men in the highest tertile of BMI (Table 7) are suggestive of effects involving tumor promotion.

Some aspects of the interaction between smoking and BMI might be explained by the effects of those factors on testosterone levels. Testosterone has been shown to act as a promoter of prostate cancer in an animal model, 25 and higher total testosterone levels in men are associated with increased risks of prostate cancer. 26 Small changes in hormone levels might cause large changes in risk. 27

Increasing BMI is associated with decreasing levels of total serum testosterone. 28,29 This relation might explain why the risk of prostate cancer was lower among nonsmokers with high BMI than among nonsmokers with low BMI (Table 6).

The association of cigarette smoking with higher levels of total serum testosterone 28,30,31 might explain the increased risk of prostate cancer associated with cigarette smoking among men with high BMI relative to not smoking among men with similar BMI (Table 7). The effects of the increase in total serum testosterone associated with smoking on prostate cancer risk might be more apparent among obese men who tend to have lower testosterone levels (Table 7) than among men with lower BMI who would tend to have higher testosterone levels (Table 6). 28,29

Nevertheless, attempts to explain prostate cancer incidence on the basis of serum hormone levels should be considered cautiously, because circulating hormone levels do not reflect intraprostatic hormonal activity. 27 Furthermore, the effects of smoking and obesity on serum testosterone may be counteracted by their effects on sex hormone binding globulin (SHBG): smoking is associated with increased levels of SHBG, and obesity is associated with decreased levels of SHBG. 28,29 SHBG binds a portion of circulating testosterone; unbound testosterone is considered to be the biologically active form. Thus, increasing levels of SHBG are associated with decreasing risks of prostate cancer, when levels of other hormones are controlled for. 26 Consequently, the increase in total serum testosterone associated with smoking that may increase risk is accompanied by an increase in SHBG that may decrease risk. The net effect of these opposing processes on prostate cancer risk may depend on BMI (Table 6). It is possible that the point estimates of the ORs in Table 6 that suggest a protective effect of smoking among men with low BMI could reflect such a net effect.

It has been suggested that smoking might also increase prostate cancer risk by reducing endogenous estrogen levels. The evidence cited for the effects of smoking on estrogen levels, however, came from studies involving women, 32–34 and Dai et al31 showed that smoking in men was not correlated with total serum estradiol or serum estrone.

We found two lines of evidence for an effect of tobacco smoking on the risk of prostate cancer. Cigar and pipe smoking seemed to be associated with the development of prostate cancer. In addition, among men with high BMI, cigarette smoking was associated with an increased risk of prostate cancer.

Back to Top | Article Outline


1. Yatani R, Chigusa I, Akazaki K, Stemmermann GN, Welsh RA, Correa P. Geographic pathology of latent prostatic carcinoma. Int J Cancer 1982; 29: 611–616.
2. Giovannucci E. Epidemiologic characteristics of prostate cancer. Cancer 1995; 75 (suppl 7): 1766–1777.
3. Colditz G. Consensus conference: smoking and prostate cancer. Cancer Causes Control 1996; 7: 560–562.
4. Giovannucci E, Rimm E, Ascherio A, Colditz GA, Spiegelman D, Stampfer MJ, Willett WC. Smoking and risk of total and fatal prostate cancer in United States health professionals. Cancer Epidemiol Biomarkers Prev 1999; 8: 277–282.
5. Pitot HC. Stages in neoplastic development. In: Schottenfeld D, Fraumeni JF, eds. Cancer Epidemiology and Prevention. 2nd ed. New York: Oxford University Press, 1996; 65–79.
6. Andersson SO, Wolk A, Bergström R, Adami HO, Engholm G, Englund A, Nyrén O. Body size and prostate cancer: a 20-year follow-up study among 135,006 Swedish construction workers. J Natl Cancer Inst 1997; 89: 385–389.
7. Siemiatycki J. Risk factors for cancer in the workplace. Boca Raton, FL: CRC Press, 1991.
8. Siemiatycki J, Krewski D, Franco E, Kaiserman M. Associations between cigarette smoking and each of 21 types of cancer: a multi-site case-control study. Int J Epidemiol 1995; 24: 504–514.
9. Richardson L. Case-control study methods: response rates, respondent characteristics and nonresponse bias (Master’s thesis). Montréal: McGill University, 1985; 97.
10. Skinner HA, Holt S, Israel Y. Early identification of alcohol abuse. 1. Critical issues and psychosocial indicators for a composite index. Can Med Assoc J 1981; 124: 1141–1152.
11. Breslow NE, Day NE. Statistical Methods in Cancer Research. vol. 1. The Analysis of Case-Control Studies. IARC Scientific Pub. No. 32. Lyon: International Agency for Research on Cancer, 1980; 192–246.
12. SAS Institute. SAS/STAT Software: Changes and Enhancements through Release 6.12. Cary, NC: SAS Institute, 1997; 871–948.
13. Lumey LH. Prostate cancer and smoking: a review of case-control and cohort studies. Prostate 1996; 29: 249–260.
14. Nomura AMY, Kolonel LN. Prostate cancer: a current perspective. Epidemiol Rev 1991; 13: 200–227.
15. Ross RK, Henderson BE. Do diet and androgens alter prostate cancer risk via a common etiologic pathway? J Natl Cancer Inst 1994; 86: 252–254.
16. Ross RK, Schottenfeld D. Prostate cancer. In: Schottenfeld D, Fraumeni JF, eds. Cancer Epidemiology and Prevention. 2nd ed. New York: Oxford University Press, 1996; 1180–1206.
17. Kolonel LN, Nomura AMY, Cooney RV. Dietary fat and prostate cancer: current status. J Natl Cancer Inst 1999; 91: 414–428.
18. Stellman SD. Chairman’s remarks: selection, follow-up, and analysis in prospective studies. Natl Cancer Inst Monogr 1985; 67: 145–147.
19. Tillotson JL, Gorder DD, Kassim N. Nutrition data collection in the Multiple Risk Factor Intervention Trial (MRFIT). J Am Diet Assoc 1981; 78: 235–240.
20. Hofstetter A, Schutz Y, Jéquier E, Wahren J. Increased 24-hour energy expenditure in cigarette smokers. N Engl J Med 1986; 314: 79–82.
21. Cerhan JR, Torner JC, Lynch CF, Rubenstein LM, Lemke JH, Cohen MB, Lubaroff DM, Wallace RB. Association of smoking, body mass, and physical activity with risk of prostate cancer in the Iowa 65+ Rural Health Study (United States). Cancer Causes Control 1997; 8: 229–238.
22. Hartman TJ, Albanes D, Rautalahti M, Tangrea JA, Virtamo J, Stolzenberg R, Taylor PR. Physical activity and prostate cancer in the alpha-tocopherol, beta-carotene (ATBC) cancer prevention study (Finland). Cancer Causes Control 1998; 9: 11–18.
23. Giovannucci E, Leitzmann M, Spiegelman D, Rimm EB, Colditz GA, Stampfer MJ, Willett WC. A prospective study of physical activity and prostate cancer in male health professionals. Cancer Res 1998; 58: 5117–5122.
24. Liu S, Lee IM, Linson P, Ajani U, Buring JE, Hennekens CH. A prospective study of physical activity and risk of prostate cancer in US physicians. Int J Epidemiol 2000; 29: 29–35.
25. Pollard M, Luckert PH. Autochthonous prostate adenocarcinomas in Lobund-Wistar rats: a model system. Prostate 1987; 11: 219–227.
26. Gann PH, Hennekens CH, Ma J, Longcope C, Stampfer MJ. Prospective study of sex hormone levels and risk of prostate cancer. J Natl Cancer Inst 1996; 88: 1118–1126.
27. Hsing AW. Hormones and prostate cancer: where do we go from here? J Natl Cancer Inst 1996; 88: 1093–1095.
28. Field AE, Colditz GA, Willett WC, Longcope C, McKinlay JB. The relation of smoking, age, relative weight, and dietary intake to serum adrenal steroids, sex hormones, and sex hormone-binding globulin in middle-aged men. J Clin Endocrinol Metab 1994; 79: 1310–1316.
29. Dai WS, Kuller LH, LaPorte RE, Gutai JP, Falvo-Gerard L, Caggiula A. The epidemiology of plasma testosterone levels in middle-aged men. Am J Epidemiol 1981; 114: 804–816.
30. Deslypere JP, Vermeulen A. Leydig cell function in normal men: effect of age, life-style, residence, diet, and activity. J Clin Endocrinol Metab 1984; 59: 955–962.
31. Dai WS, Gutai JP, Kuller LH, Cauley JA. Cigarette smoking and serum sex hormones in men. Am J Epidemiol 1988; 128: 796–805.
32. Baron JA. Smoking and estrogen-related disease. Am J Epidemiol 1984; 119: 9–22.
33. MacMahon B, Trichopoulos D, Cole P, Brown J. Cigarette smoking and urinary estrogens. N Engl J Med 1982; 307: 1062–1065.
34. Michnovicz JJ, Hershcopf RJ, Naganuma H, Bradlow HL, Fishman J. Increased 2-hydroxylation of estradiol as a possible mechanism for the anti-estrogenic effect of cigarette smoking. N Engl J Med 1986; 315: 1305–1309.

prostatic neoplasms; smoking; body mass index; case-control studies

© 2001 Lippincott Williams & Wilkins, Inc.