The Prostate 17:189-206 (1990)
Epidemiology of Prostatic Cancer: A Case-Control Study Shirley M. Fincham, Gerry 8. Hill, John Hanson, and Chandrani Wijayasinghe Division of Epidemiology and Preventive Oncology, Alberta Cancer Board (S.M.F., J.H.), Professions and Occupations Bureau, Solicitor General (C. W.), Edmonton, Alberta, and Department of Epidemiology and Biostatistics, McGill University, Montreal, Quebec, (G.6.H.), Canada A population-based case-control study of prostatic cancer in Alberta was undertaken to determine the risk factors associated with the disease. Cases were 382 newly diagnosed prostatic cancer patients and 625 controls, group-matched to the anticipated age distribution of the cases, chosen at random from the health insurance roster. Subjects were interviewed in their homes by using a pre-tested questionnaire including questions related to ethnic group, education, puberty, marital history, family history, residence, water supply, smoking, and diet. Factors significantly related to the risk of developing prostatic cancer included ethnic group (British high, Ukrainian low), education (elementary high, university low), age at first marriage (early high, late low), family history (high risk for those with relatives with prostatic cancer), and increased masculinity among the children of cases. The results with respect to smoking, occupation, medical history, birthplace, residence, water supply, and diet were generally negative.
Key words: risk factors, ethnicity, fertility, education, family history
INTRODUCTION Cancer of the prostate is one of the more frequent malignant neoplasms in men over 50 and is increasing in incidence throughout most of the world. In a recent review of 46 cancer registries [l], increasing incidence was noted in 39. Incidence tends to be high in North America and Western Europe and low in Eastern Europe, Africa, and Asia. Although the incidence of prostatic cancer is low in African blacks, rates in the United States are highest among blacks and mortality from prostate cancer in the United States has risen more sharply among non-whites than among whites. It has been shown that this is a cohort phenomenon with a higher risk for non-whites born around the turn of the century [2]. A number of investigators have examined risk factors associated with prostatic cancer, but the results have been inconsistent or inconclusive. A genetic etiology has
Received for publication February 22, 1990; accepted June 15, 1990. Address reprint requests to Shirley M. Fincham, Division of Epidemiology and Preventive Oncology, 6th Floor, 9707-1 10 Street, Edmonton, Alberta T5K 2L9, Canada. 0 1990 Wiley-Liss, Inc.
190
Fincham et al.
been proposed, but no specific Mendelian syndromes have been described in relation to prostate cancer. Increased cancer incidence has been found in relatives of prostatic cancer patients in several studies [3,4], but not in all [5]. Although there is no reason to believe that the prostate is immune to the carcinogenic effects of radiation, there is no evidence to show that this is an important factor in man. Studies of atom bomb survivors [6] have shown no excess risk of prostate cancer. There is no evidence that radiologists or others occupationally exposed to radiation have increased risk of prostate cancer [7,8]. Some studies [9,10] have examined such factors as puberty, body size and shape, and the distribution of hair. Although some differences have been reported, the results are in general negative or inconclusive. Cross-sectional studies [ 11,121 of mortality based on vital statistics show that prostate cancer is one of the few diseases for which the rates are higher for married men. However, this has not always been found in case-control studies [9,13,14]. Although case-control studies 114-181 have not been completely consistent, the findings in general suggest that prostatic cancer patients have experienced a greater sexual drive than controls, with more sexual partners, earlier marriage, and greater sexual activity within marriage. An increase in fertility among prostate cancer patients has been reported in a number of studies [10,13,17] but the differences were statistically significant in only one. A similar conflict exists in relation to the association of benign prostatic hypertrophy and prostate cancer. A positive association has been found in some studies 15,191 but a large cohort study [20] found no association. Three viruses have been suggested as possible etiologic agents-herpes type 2, simian virus 40, and cytomegalic virus. The findings in relation to marital and sexual history, including an excess of venereal disease in some studies [5,15-171, would favour the hypothesis, but studies of the association between prostate cancer and cancer of the cervix in wives have been negative or inconclusive [5,2 I]. Interest in chemical carcinogens has focused chiefly on trace metals, especially cadmium and zinc. Several occupational and environmental studies have implicated cadmium indirectly [22,23]. Again, the studies are not conclusive. In common with several other cancer sites, thtre is much interest in the possible role of diet. As with breast and colorectal cancer, geographic studies [9,23,24,25] have shown a positive correlation between prostate cancer and fat consumption and this has also been found in case control studies [26,27]. A study in Japan found a significant reduction in mortality from prostate cancer in men who ate green and yellow vegetables daily [28], but in a case-control study in the United States a positive association was found with intake of retinoids and vitamin C 1271. To clarify these findings, we did a three-year study of risk factors in new cases diagnosed with primary prostatic cancer in Alberta, compared to risk factor prevalence in an age-stratified random sample of controls selected from the general Alberta population. Our data provide a comparison with previous work and suggest some new areas for investigation. MATERIALS AND METHODS Subject Selection and Data Collection Cases were identified through the Alberta Cancer Registry, a population-based cancer registry in operation since 1941, which includes virtually all cancer cases
Epidemiology of Prostatic Cancer
191
diagnosed in the province. Cases were histologically confirmed, diagnosed between April 1981 and September 1983, and 45 years of age and older. Controls, groupmatched to the anticipated age distribution of cases, were selected at random from the roles of the Alberta Health Care Insurance Plan, which provides medical care insurance for all residents of the province. Subjects were interviewed in their homes by trained interviewers, using pretested questionnaires. In addition to identifying information on both the subject and his wife, interviewers asked questions about ethnicity , education, marital status, occupational history, personal and family medical history, tobacco and alcohol consumption, age at puberty, and physical constitution. Subjects were also asked about sexual behaviour patterns, including whether their sexual activities had been limited at any time for any reason. A detailed dietary history was obtained by using an instrument which assesses frequency and amount of food consumed over two 2-month periods separated by six months. The diet history was developed by the Epidemiology Unit of the National Cancer Institute (Canada) and was described in detail in a previous publication [29]. To help respondents estimate amount of food consumed, food models were constructed from a variety of materials and presented to the subject during the interview in a standard order. Prevalence of cancer in wives of respondents was determined by matching both mamed and maiden names against the Alberta Cancer Registry. Cancer site was recorded for matches; if the name did not appear, the wife was considered to be cancer-free. Analysis The questionnaires were coded, checked, and computerized. Contingency tables were produced by using SPSS [30]. Relative risks for risk factors for cancer of the prostate were determined by using the odds ratios as a measure of relative risk, with chi-square tests for trend where appropriate [31]. The method tests for homogeneity of the relative risks, rather than homogeneity over age strata. Cox’s logistic regression model [32] was used to examine the combined influence of significant odds ratios. RESULTS
Results are presented for 382 cases and 625 controls. As the age distributions were so similar, the following analyses are presented without stratification by age, but statistically significant risks were confirmed by age-adjusted analyses. Demographic Variables and Cigarette Use Table I shows relative risks for educational status, ethnicity, and cigarette smoking. There was an inverse relationship between highest level of education obtained and risk, those with university education having a significantly reduced risk compared to those with an elementary education. The downward trend in relative risk with increasing educational attainment was also significant (X2 = 4.22, P < 0.05). The risk of prostate cancer was significantly lower for men of Ukrainian descent compared with those of British origin (Table I). Those of German descent were also
192
Fincham et al.
TABLE 1. Relative Risk of Prostatic Cancer bv Selected Characteristics
Variable Highest level of education Elema Sec . Tech. Univ. Total Homogeneity chi-square = 5.46 Trend chi-square = 4.22* Ethnicity British" French German Ukrainian Other Total Homogeneity chi-square = 10.23 * Cigarette smoking Neve?
Ex Current Total Homogeneity chi-square = I .38 Trend chi-square = 0.65 No. of cigarettes smoked 1-10 1 1-20 21-30 31 + Total Homogeneity chi-square = 7.43 Trend chi-sauare = 2.55
Relative risk
95% confidence interval
cases
Controls
197 97 46 33 373
303 145 87 81 616
200 16 55 31 74 376
279 20 93 79 149 620
1.12 0.83 0.55 0.69
106
153 317 155 625
0.83 0.87
0.61-1.13 0.61-1.24
0.82 0.88 1 .OO 0.57
0.57-1.18 0.60-1.29 0.68-1.48 0.36-0.88
183 93 382
86 71 73 37 373
151 117 105
95 62 1
1
1.03 0.81 0.63
0.75-1.41 0.55-1.21 0.40-0.98
1
0.56-2.21 0.56-1.21 0.35-0.86 0.50-0.97
1
'Reference group. *0.01
< P < 0.05.
at reduced risk but the difference did not reach significance, as was the case for the slightly elevated risk for those of French descent. The risk for cigarette smokers was lower than that for those who had never smoked, but the reductions were not significant. Nor was there any evidence of a trend in risk with the number of cigarettes smoked by smokers and ex-smokers (Table I). Marital Status and Sexual Behaviour
Table I1 lists relative risks associated with marital status and the variables related to sexual behaviour. Analysis by marital status shows that those who had never mamed were at almost twice the risk of those who were married, or were previously married. Al-
Epidemiology of Prostatic Cancer
193
TABLE 11. Relative Risks for Prostatic Cancer by Marital Status and Sexual Behaviour
Variable Marital status Marrieda Widowed/single/divorced Never married Total Homogeneity chi-square = 5.14 No. of never married" Marriages 1
2+ Total Homogeneity chi-square = 5.03 Trend chi-square = 1.37 Age at first marriage
Epidemiology of Prostatic Cancer: A Case-Control Study Shirley M. Fincham, Gerry 8. Hill, John Hanson, and Chandrani Wijayasinghe Division of Epidemiology and Preventive Oncology, Alberta Cancer Board (S.M.F., J.H.), Professions and Occupations Bureau, Solicitor General (C. W.), Edmonton, Alberta, and Department of Epidemiology and Biostatistics, McGill University, Montreal, Quebec, (G.6.H.), Canada A population-based case-control study of prostatic cancer in Alberta was undertaken to determine the risk factors associated with the disease. Cases were 382 newly diagnosed prostatic cancer patients and 625 controls, group-matched to the anticipated age distribution of the cases, chosen at random from the health insurance roster. Subjects were interviewed in their homes by using a pre-tested questionnaire including questions related to ethnic group, education, puberty, marital history, family history, residence, water supply, smoking, and diet. Factors significantly related to the risk of developing prostatic cancer included ethnic group (British high, Ukrainian low), education (elementary high, university low), age at first marriage (early high, late low), family history (high risk for those with relatives with prostatic cancer), and increased masculinity among the children of cases. The results with respect to smoking, occupation, medical history, birthplace, residence, water supply, and diet were generally negative.
Key words: risk factors, ethnicity, fertility, education, family history
INTRODUCTION Cancer of the prostate is one of the more frequent malignant neoplasms in men over 50 and is increasing in incidence throughout most of the world. In a recent review of 46 cancer registries [l], increasing incidence was noted in 39. Incidence tends to be high in North America and Western Europe and low in Eastern Europe, Africa, and Asia. Although the incidence of prostatic cancer is low in African blacks, rates in the United States are highest among blacks and mortality from prostate cancer in the United States has risen more sharply among non-whites than among whites. It has been shown that this is a cohort phenomenon with a higher risk for non-whites born around the turn of the century [2]. A number of investigators have examined risk factors associated with prostatic cancer, but the results have been inconsistent or inconclusive. A genetic etiology has
Received for publication February 22, 1990; accepted June 15, 1990. Address reprint requests to Shirley M. Fincham, Division of Epidemiology and Preventive Oncology, 6th Floor, 9707-1 10 Street, Edmonton, Alberta T5K 2L9, Canada. 0 1990 Wiley-Liss, Inc.
190
Fincham et al.
been proposed, but no specific Mendelian syndromes have been described in relation to prostate cancer. Increased cancer incidence has been found in relatives of prostatic cancer patients in several studies [3,4], but not in all [5]. Although there is no reason to believe that the prostate is immune to the carcinogenic effects of radiation, there is no evidence to show that this is an important factor in man. Studies of atom bomb survivors [6] have shown no excess risk of prostate cancer. There is no evidence that radiologists or others occupationally exposed to radiation have increased risk of prostate cancer [7,8]. Some studies [9,10] have examined such factors as puberty, body size and shape, and the distribution of hair. Although some differences have been reported, the results are in general negative or inconclusive. Cross-sectional studies [ 11,121 of mortality based on vital statistics show that prostate cancer is one of the few diseases for which the rates are higher for married men. However, this has not always been found in case-control studies [9,13,14]. Although case-control studies 114-181 have not been completely consistent, the findings in general suggest that prostatic cancer patients have experienced a greater sexual drive than controls, with more sexual partners, earlier marriage, and greater sexual activity within marriage. An increase in fertility among prostate cancer patients has been reported in a number of studies [10,13,17] but the differences were statistically significant in only one. A similar conflict exists in relation to the association of benign prostatic hypertrophy and prostate cancer. A positive association has been found in some studies 15,191 but a large cohort study [20] found no association. Three viruses have been suggested as possible etiologic agents-herpes type 2, simian virus 40, and cytomegalic virus. The findings in relation to marital and sexual history, including an excess of venereal disease in some studies [5,15-171, would favour the hypothesis, but studies of the association between prostate cancer and cancer of the cervix in wives have been negative or inconclusive [5,2 I]. Interest in chemical carcinogens has focused chiefly on trace metals, especially cadmium and zinc. Several occupational and environmental studies have implicated cadmium indirectly [22,23]. Again, the studies are not conclusive. In common with several other cancer sites, thtre is much interest in the possible role of diet. As with breast and colorectal cancer, geographic studies [9,23,24,25] have shown a positive correlation between prostate cancer and fat consumption and this has also been found in case control studies [26,27]. A study in Japan found a significant reduction in mortality from prostate cancer in men who ate green and yellow vegetables daily [28], but in a case-control study in the United States a positive association was found with intake of retinoids and vitamin C 1271. To clarify these findings, we did a three-year study of risk factors in new cases diagnosed with primary prostatic cancer in Alberta, compared to risk factor prevalence in an age-stratified random sample of controls selected from the general Alberta population. Our data provide a comparison with previous work and suggest some new areas for investigation. MATERIALS AND METHODS Subject Selection and Data Collection Cases were identified through the Alberta Cancer Registry, a population-based cancer registry in operation since 1941, which includes virtually all cancer cases
Epidemiology of Prostatic Cancer
191
diagnosed in the province. Cases were histologically confirmed, diagnosed between April 1981 and September 1983, and 45 years of age and older. Controls, groupmatched to the anticipated age distribution of cases, were selected at random from the roles of the Alberta Health Care Insurance Plan, which provides medical care insurance for all residents of the province. Subjects were interviewed in their homes by trained interviewers, using pretested questionnaires. In addition to identifying information on both the subject and his wife, interviewers asked questions about ethnicity , education, marital status, occupational history, personal and family medical history, tobacco and alcohol consumption, age at puberty, and physical constitution. Subjects were also asked about sexual behaviour patterns, including whether their sexual activities had been limited at any time for any reason. A detailed dietary history was obtained by using an instrument which assesses frequency and amount of food consumed over two 2-month periods separated by six months. The diet history was developed by the Epidemiology Unit of the National Cancer Institute (Canada) and was described in detail in a previous publication [29]. To help respondents estimate amount of food consumed, food models were constructed from a variety of materials and presented to the subject during the interview in a standard order. Prevalence of cancer in wives of respondents was determined by matching both mamed and maiden names against the Alberta Cancer Registry. Cancer site was recorded for matches; if the name did not appear, the wife was considered to be cancer-free. Analysis The questionnaires were coded, checked, and computerized. Contingency tables were produced by using SPSS [30]. Relative risks for risk factors for cancer of the prostate were determined by using the odds ratios as a measure of relative risk, with chi-square tests for trend where appropriate [31]. The method tests for homogeneity of the relative risks, rather than homogeneity over age strata. Cox’s logistic regression model [32] was used to examine the combined influence of significant odds ratios. RESULTS
Results are presented for 382 cases and 625 controls. As the age distributions were so similar, the following analyses are presented without stratification by age, but statistically significant risks were confirmed by age-adjusted analyses. Demographic Variables and Cigarette Use Table I shows relative risks for educational status, ethnicity, and cigarette smoking. There was an inverse relationship between highest level of education obtained and risk, those with university education having a significantly reduced risk compared to those with an elementary education. The downward trend in relative risk with increasing educational attainment was also significant (X2 = 4.22, P < 0.05). The risk of prostate cancer was significantly lower for men of Ukrainian descent compared with those of British origin (Table I). Those of German descent were also
192
Fincham et al.
TABLE 1. Relative Risk of Prostatic Cancer bv Selected Characteristics
Variable Highest level of education Elema Sec . Tech. Univ. Total Homogeneity chi-square = 5.46 Trend chi-square = 4.22* Ethnicity British" French German Ukrainian Other Total Homogeneity chi-square = 10.23 * Cigarette smoking Neve?
Ex Current Total Homogeneity chi-square = I .38 Trend chi-square = 0.65 No. of cigarettes smoked 1-10 1 1-20 21-30 31 + Total Homogeneity chi-square = 7.43 Trend chi-sauare = 2.55
Relative risk
95% confidence interval
cases
Controls
197 97 46 33 373
303 145 87 81 616
200 16 55 31 74 376
279 20 93 79 149 620
1.12 0.83 0.55 0.69
106
153 317 155 625
0.83 0.87
0.61-1.13 0.61-1.24
0.82 0.88 1 .OO 0.57
0.57-1.18 0.60-1.29 0.68-1.48 0.36-0.88
183 93 382
86 71 73 37 373
151 117 105
95 62 1
1
1.03 0.81 0.63
0.75-1.41 0.55-1.21 0.40-0.98
1
0.56-2.21 0.56-1.21 0.35-0.86 0.50-0.97
1
'Reference group. *0.01
< P < 0.05.
at reduced risk but the difference did not reach significance, as was the case for the slightly elevated risk for those of French descent. The risk for cigarette smokers was lower than that for those who had never smoked, but the reductions were not significant. Nor was there any evidence of a trend in risk with the number of cigarettes smoked by smokers and ex-smokers (Table I). Marital Status and Sexual Behaviour
Table I1 lists relative risks associated with marital status and the variables related to sexual behaviour. Analysis by marital status shows that those who had never mamed were at almost twice the risk of those who were married, or were previously married. Al-
Epidemiology of Prostatic Cancer
193
TABLE 11. Relative Risks for Prostatic Cancer by Marital Status and Sexual Behaviour
Variable Marital status Marrieda Widowed/single/divorced Never married Total Homogeneity chi-square = 5.14 No. of never married" Marriages 1
2+ Total Homogeneity chi-square = 5.03 Trend chi-square = 1.37 Age at first marriage
