Epidemiologic Reviews Copyright © 2001 by the Johns Hopkins University Bloomberg School of Public Health All rights reserved

Vol. 23, No. 1 Printed in U.S.A.

Smoking and Prostate Cancer

Kym Hickey,1 Kim-Anh Do, 2 and Adele Green 3

associations between current smoking and prostate cancer ranged in both magnitude and direction, from negative, 0.38 (72), to positive, 2.9 (71). The association between past smoking and prostate cancer also ranged from negative, 0.6 (35, 91), to positive, 1.9 (51). The quality of studies varied greatly, with scores ranging from 2.5, out of 22 (80), to 19, out of 22 (34). There was little evidence of publication bias, as seen from figure 1. Most of the prospective cohort studies and all of the nested case-control studies that used incident cases as the outcome found no association between current smoking and prostate cancer (34,35,37,39^2,44,45, 56-59). However, a positive association was observed in a cohort of elderly rural residents of Iowa (36), in a cohort of health plan members in California (38), and in a cohort of Swedish construction workers (43). In the Iowa study (36), the association with current smoking was more pronounced with regional and distant prostate cancer (relative risk (RR) = 8.7; 95 percent confidence interval (CI): 1.9,40) than localized disease (RR = 2.5; 95 percent CI: 0.9,7.1). The Iowa study (36) and the Swedish construction workers study (43) found evidence of a significant dose-response relation between current smoking and risk of incident prostate cancer. The majority of the prospective cohort studies that used death from prostate cancer as the outcome, in contrast, found a positive association between current smoking and prostate cancer (43, 46, 47, 49, 51-53, 60). The strength of the association observed between current smoking and fatal prostate cancer was weak—on the order of approximately a 30 percent increase. While four of the 12 mortality cohort studies (48, 50, 54, 55) did not find a positive relation, the upper limit of the confidence intervals of the relative risk ranged from 1.2 to 1.5. The US veterans cohort found a significant dose-response relation between current smoking and fatal prostate cancer (49). The Health Professionals Foliow-Up Study found that men who had smoked 15 or more pack-years of cigarettes within the preceding 10 years were at a higher risk of distant metastatic prostate cancer (RR = 1.81; 95 percent CI: 1.05, 3.11) and fatal prostate cancer (RR = 2.06; 95 percent CI: 1.08, 3.90) relative to nonsmokers (34). This study also found a significant doseresponse relation between smoking over the prior 10 years and distant metastatic and fatal prostate cancer (34). Ten of 15 population-based case-control studies (66, 70, 73, 74, 81-83, 89, 94, 96) and 12 of 16 hospital-based casecontrol studies (62, 69, 77, 80, 84-88, 91, 93, 95) produced essentially null results with regard to current smoking and prostate cancer. Five other case-control studies used hospi-

INTRODUCTION

The relation between smoking and many cancers is established but its role in prostate cancer, the most common cancer in US males, is not clear. Given the high prevalence of smoking and the high prostate cancer morbidity, there are substantial public health implications if smoking is related to prostate cancer, and, hence, a review of this topic is warranted. METHODS

A Medline search was conducted and covered the period from 1966 to March 2000. The bibliographies of identified papers were reviewed, and three papers presented at a conference were ascertained. Only analytical studies which contained a specific prostate cancer-smoking estimate (or provided enough data to allow calculation of such an estimate) and were written in English were included in this review. Table 1 lists those studies not included (1-13). Each of the studies included was scored for quality by a reviewer, and 58 percent of the studies were independently scored by a second reviewer. The scoring system developed incorporated the main methodological issues of study design, selection issues, measurement of prostate cancer and smoking status, control of the five most likely confounding factors identified (age (14, 15, 16), race (14, 17, 18), family history of prostate cancer (17, 19-25), dietary fat (26), history of vasectomy (27-33)), and statistical methods. There was excellent agreement between reviewers in the scoring of the case-control studies and reasonable agreement in the scoring of cohort studies. EPIDEMIOLOGIC STUDIES

Tables 2-5 describe characteristics of the 23 prospective cohort studies (34—55), five nested case-control studies (56-60), one retrospective cohort study (61), and 36 casecontrol studies (62-97) included in this review. The reported Received for publication July 31, 2000, and accepted for publication November 16, 2000. Abbreviations: Cl, confidence interval; DHEAS, dehydroepiandrosterone sulphate; RR, relative risk. 1 Repatriation Medical Authority, Brisbane, Australia. 2 Department of Biostatistics, University of Texas MD Anderson Cancer Center, Houston, TX. 3 Queensland Institute of Medical Research, Brisbane, Australia. Reprint requests to Dr. Kym Hickey, Repatriation Medical Authority, GPO Box 1014, Brisbane, Queensland 4001, Australia (email: [email protected]).

115

116

Hickeyetal.

TABLE 1. Studies excluded from review and reason for exclusion Study description (reference no.)

Weir and Dunn (1), 1970, United States, prospective cohort Armenian et at. (2), 1975, United States, case-control study Williams and Horm (3), 1977, United States, cross-sectional study Kolonel and Winkelstein (4), 1977, United States, case-control study Niijima and Koiso (5), 1980, Japan, case-control study Jackson et al. (6), 1980, United States and Nigeria, case-control study Wigle et al. (7), 1980, Canada, case-control study Newell et al. (8), 1989, United States, case-control study LeMarchand et al. (9), 1991, Hawaii, case-control study Van Der Gulden et al. (10), 1992, The Netherlands, case-control study Pawlega et al. (11), 1996, Poland, case-control study Tulinius et al. (12), 1997, Iceland, prospective cohort Deneo-Pellegrini et al. (13), 1999, Uruguay, case-control study

Data provided on smoking and prostate cancer

Reason for exclusion

Relative risk = 0.78

95% confidence interval not provided*

No major difference between case and control groups regarding smoking Relative risk = 0.81 (calculated from data provided)

Estimate of effect not provided* 95% confidence interval not provided*

Odds ratio = 1.1

95% confidence interval not provided*

Almost the same proportions in both study groups were smokers When cases and controls were compared no apparent trend was detected for smoking Odds ratio = 0.9

Estimate of effect not provided*

Cases and controls did not differ with regard to smoking No statistically significant difference between cases and controls in smoking habits Cases and controls did not differ in smoking habits Cigarette smoking did not significantly influence prostate cancer Smoking was not a significant risk factor for prostate cancer Tobacco consumption not associated with prostate cancer risk

Estimate of effect not provided*

95% confidence interval not provided* Estimate of effect not provided* Estimate of effect not provided*

Estimate of effect not provided* Estimate of effect not provided* Estimate of effect not provided* Estimate of effect not provided*

• And unable to be calculated from the data provided by the study.

tal cases and population controls, and four of these studies produced null findings (67, 68, 92, 97). These studies suffered from several methodological shortcomings. Differentia] measurement error may have been present in all mortality cohorts stemming from the use of cause of death as the study endpoint. Smokers are more likely to die, especially from tobacco-related diseases (98). Hence, individuals with prostate cancer who smoked would be more likely to die than individuals with prostate cancer who did not smoke, and if prostate cancer was recorded as the underlying cause of death on the death certificate when it was not, more thorough ascertainment of prostate cancer in smokers would occur (46). The Health Professionals Follow-up Study (34) conducted an analysis among men diagnosed with prostate cancer and showed that the presence of smoking-related comorbidities predicted death from other causes (RR = 1.63) but did not predict death attributed to prostate cancer (RR = 1.07). Surveillance bias may have been partly responsible for the positive association reported in two of the incident cohort studies (36, 43) as smokers may have had more contact with medical practitioners due to other smoking-related diseases that required treatment. Investigators of the Health Professionals Follow-up Study (34) also looked at the possibility that a lower frequency of screening at baseline among smokers may have accounted for the elevated mortality from prostate cancer found in their

study (through a delay in diagnosis and treatment). In an analysis limited to men who had reported a negative digital rectal examination in 1986-1988, the associations between smoking and distant metastatic and fatal prostate cancer from 1988 to 1994 became even stronger. A recent study found that nicotine dependence was not related to prostate cancer screening (99). Uncontrolled confounding by dietary fat was a threat in most studies as only four adjusted for it (34,46, 66, 91), and another four studies did not find an association between dietary fat and prostate cancer (35, 51, 89, 96). Assuming dietary fat is a risk factor for prostate cancer, dietary fat could confound the relation between smoking and prostate cancer (100, 101), since numerous cross-sectional studies have found that smokers are more likely to consume a diet high in fat (102-112). No study appeared to have considered the effects of treatment received by prostate cancer cases. If smokers received less aggressive therapy than nonsmokers, then this may partly account for the positive association observed with fatal prostate cancer. Four studies have followed up cases of prostate cancer and found a lower survival rate among prostate cancer cases who smoked compared with nonsmokers (34, 113-115), although a hospital-based study in Japan differed (93). One interpretation of the findings relating relatively recent smoking to advanced incident prostate cancer or fatal Epidemiol Rev Vol. 23, No. 1, 2001

.

Study characteristics of prospective cohorts/nested case-control studies investigating incident prostate cancer

Study ference no.)

Location, years'

Norway, 1977-1983

16-56

69/24,051

n et al. (36), 1997

United States, 1981-1982

65-101

71/1,050

t al. (37), 1989

United States, 1974

225, white

180/14,000

t al. (38), 1994

United States, 1978-1985

238/43,432

rchand et al. (39), 4

United States, 1975-1980

39-87, white and black 245, multiethnic (Hawaii)

and et al. (40), 6 son et al. (41), 9 and Lund (42), 4 et al. (43), 1996

Norway, 1964-1965

38-72

707/11,863

46-68, Japanese

174/7,999

Norway, 1972-1978

35-50

220/43,685

Sweden, 1971-1975

60

2,368/135,006

Finland, 1962

30-74

209/4,601

Finland, 1972,1977

15-74

109/11,373

Retired, white

138/5,106

42-82

406/9,152

50-84, white

54/1,776

215, white

166/16,481

United States, 1916-1950

College students

243/29,000

United States, 1986-1989

40-86

81/1,177

Sweden, 1967,1970 United States, 1972-1974 Finland, 1966-1972

Age

17

Age, race

16

Age, race

16

Age, race

16

Age

16

Age, race

15

Age

15

1.04; 0.85, 1.27 1.00;0.92, 1.08 0.56; 0.36, 0.83 0.6; 0.3, 1.1 2.2; 1.2,4.4 1.2; 0.7, 2.1 0.49;0.16,1.57 1.24; 0.91, 1.67 1.46;1.07,1.94 1.1;0.8, 1.5 Quartiles in cigar day: 1) 1.0 2)0.9;0.6, 1 3) 1.0; 0.7, 4) 1.0; 0.6, 1.1; 0.9, 1.3 0.9;0.7, 1.1 0.87:0.61,1.23 0.89:0.61, 1.29 1.08:0.90,1.30

Age

15

Age?

11

Age?

11

Age

10

Age

15

Age

14

1.11:1.01,1.23 1.09; 0.98, 1.22 1.11:0.90,1.36 0.93; 0.68, 1.24 0.82; 0.57, 1.14 1.07; 0.80, 1.40 0.9;0.40,1.73 0.8;0.62, 1.01 1.00:0.71,1.39 0.91;0.68, 1.21 1.3; 0.61, 2.79

Current smoker

Age, race

14

0.76;0.51,1.14

Ever cigarettes

Age

13

1.1;1.0,1.3

Current cigarettes Past cigarettes

Age

12

1.79:1.01,2.94 1.3;0.8,2.2

Current cigarettes Past cigarettes Current cigarettes Past cigarettes Ever cigarettes

1,369/47,781

erg et al. (58), 6H pson et al. (57), 9H a et al. (59), 911 more et al. (56), 4D et al. (61), 1999H

19

Current cigarettes Past cigarettes Current cigarettes Past cigarettes Current cigarettes Past cigarettes Current cigarettes Past cigarettes Current cigarettes Past cigarettes Ever cigarettes

40-75

United States, 1981-1985

Age, race, family history, dietary fat, vasectomy Age, dietary fat

Smoking definition!

United States, 1986

nen et al. I (44), 6 nen et al. II (44), 6 et al. (45), 1990

Relative risk; 9 confidence inter

No. of cases/ cohort size

nnucci et al. (34), 9 d et al. (35), 1997

United States, 1965-1968

Quality scorei

Age range (years), race

198/8,881

Current tobacco Past tobacco Current tobacco Past tobacco Current tobacco Past tobacco Current cigarettes Past cigarettes Current tobacco Past tobacco Current cigarettes

Confounders

17

rs to years when study subjects were enrolled. erence group: never cigarettes (34-38, 40, 41, 45, 61), never tobacco (43, 58), whole country (44), noncigarettes (57), lowest quartile (39), nonsmoker (59). es out of 22. tive risk and confidence interval calculated from data provided ((34) past cigarettes, (35) current cigarettes, (38) current cigarettes, (59) current smoker, (61) current cigarettes); confidence inte lated from data provided (44, 45, 57); relative risk expressed as risk per 10 pack-years of cigarette smoking (56); risk per 10 cigarettes (42). ted case-control study (56-59); retrospective cohort study (61).

.

Study characteristics of prospective cohorts and nested case-control studies investigating fatal prostate cancer

Study ference no.)

nucci et al. (34), 9 et al. (43), 1996

uez et al. (46), 7 ond and Horn (47), 8 and Hirayama (48), 0 et al. (49), 1991

Location, years*

Age range (years), race

No. of cases/ cohort size

Smoking definitionf

Confounders

Quality

Age, race, family history, dietary fat, vasectomy Age

19

Age, race, family history, dietary fat, vasectomy Age, race

14 13

1.58:0.81,3.10 1.28; 0.95, 1.68 1.26; 1.06, 1.50 1.03; 0.84, 1.33 1.34; 1.16, 1.56 0.99; 0.87, 1.12 1.75; 1.37,2.19

scorei

Relative risk; 9 confidence inter

United States, 1986

40-75

103/47,781

Sweden, 1971-1975

60

709/135,006

United States, 1982

£30, white and black

United States, 1952

50-69, white

134/187,783

Current cigarettes Past cigarettes Current tobacco Past tobacco Current cigarettes Past cigarettes Ever cigarettes

>40

147/122, 261

Current cigarettes

Age

13

1.1; 0.7, 1.5

31-84, white

4,607/250,000

Current cigarettes Past cigarettes Current cigarettes Past cigarettes Ever cigarettes Past cigarettes Current cigarettes Past cigarettes Current cigarettes

Age

13

Age

13

Age, race, dietary fat

12

Age

12

Age, race

12

1.18; 1.09, 1.28 1.13; 1.03, 1.24 0.99; 0.87, 1.34 0.86; 0.80, 1.12 2.0; 1.1,3.7 1.9; 1.1,3.3 1.83; 1.01,3.05 1.51:0.63,3.02 1.31; 1.13, 1.52

Ever cigarettes Current tobacco Past tobacco Current cigarettes

Age Age

11 10

Japan, 1965 United States, 1954, 1957

1,748/450,279

al. (50), 1994

Great Britain, 1951

35->85

569/34,439

etal. (51), 1990

United States, 1966

>35, white

149/17,633

l et al. (52), 1993

Norway, 1972-1978

lin et al. (53), 1996

United States, 1973-1975

ond (54), 1966 nsen et al. (55), 7 lzer et al. (60), 9H

United States, 1959-1960 Sweden, 1963 Switzerland, 1971-1973

35-49 35-57, white and black 35-84 18-69 cases 64, controls 49 (mean)

32/44,290 826/348,874 319/441,542 193/25,129 30/2,974

None

15

7

1.02; 0.81, 1.28 0.93; 0.72, 1.18 1.0; 0.72, 1.35 1.38; 0.67, 2.85

rs to years when study subjects were enrolled. rence group' never cigarettes (34, 48, 52), never tobacco (43, 46, 47, 49-51, 54, 55), non-cigarettes (53), nonsmokers (60). es out of 22. tive risk and confidence interval calculated from data provided ((34) past cigarettes, (50, 52) current and past cigarettes, (54) ever cigarettes, (55) current tobacco, (60)); confidence interval on m data provided ((47) ever cigarettes, (55) past tobacco). ed case-control study that used fatal prostate cancer.

4.

Study characteristics of population-based case-control studies

Study eference no.)

Location, years*

etal. (81), 1996

China, 1992-1995

et al. (82), 1994

United States, 1986-1989

Smoking definitionf

50-94

239/472

Current cigarettes Past cigarettes Current cigarettes Past cigarettes Current cigarettes Past cigarettes Ever cigarettes

Age, race

14.5

Age, race

13

Age

13

Age

12

Age

12

Age, race, dietary fat

12

Age

12 12

0.78; 0.63, 1.09; 0.98, 1.06; 0.66, 1.05; 0.72, 1.04; 0.62, 0.93; 0.67, 0.91; 0.77,

11

1.9; 1.2, 3.0

11

1.11; 0.86, 1.44

40-79, white/ black