Alcohol consumption and prostate cancer risk: Results from the Melbourne collaborative cohort study
Although there is little evidence to support an association between alcohol consumption and prostate cancer risk, questions remain concerning the effect on aggressive and nonaggressive tumours and the pattern and type of alcohol consumed. In a prospective cohort of 16,872 men aged 27–70 years at recruitment and followed-up from 1994 to the end of 2003, 732 incident prostate cancers were identified through the local population cancer registry, including 132 aggressive cases and 53 prostate cancer deaths. Detailed information on alcohol consumption was taken at baseline by trained interviewers using a structured questionnaire. Overall, alcohol intake was not associated with prostate cancer incidence. Compared to abstainers, men consuming 1–19 g/d of alcohol had a slightly reduced incidence of aggressive prostate cancers (hazard ratio 0.67; 95% confidence interval (CI) 0.43, 1.06) and prostate cancer mortality (hazard ratio 0.56; 95% CI 0.28, 1.14), but their risk of nonaggressive prostate cancers was close to unity (hazard ratio 1.09; 95% CI 0.85, 1.40). No significant association with pattern of drinking or type of alcoholic beverage was found. Our results show that alcohol consumption does not influence overall prostate cancer incidence but we found suggestive evidence that alcohol consumption might decrease the incidence of aggressive prostate cancer and mortality. © 2006 Wiley-Liss, Inc.
The influence of drinking alcoholic beverages has been fairly extensively investigated in epidemiological studies of various cancers, but there has been little consistent evidence of association with prostate cancer risk.
Enough evidence has been gathered to conclude that, overall, alcohol consumption is not an important risk factor for prostate cancer.1 However, the possibility that a very high consumption might increase risk cannot be ruled out. Interest in alcohol and prostate cancer has been recently renewed by 2 studies providing information on patterns of consumption2 and type of alcoholic beverage consumed.3 In particular, the first study suggested that consuming large amounts of alcohol infrequently might increase risk.2
We assessed in a prospective cohort study the relationship between the consumption of alcoholic beverages and the risk of prostate cancer, overall and separately for nonaggressive, aggressive and fatal prostate cancers.
The Melbourne Collaborative Cohort Study (MCCS) is a prospective cohort study of 41,528 people (17,049 men) aged between 27 and 75 years at baseline (99.3% aged 40–69).4 Recruitment occurred between 1990 and 1994 in the Melbourne metropolitan area. Subjects were recruited via the Electoral Rolls (registration to vote is compulsory for adults in Australia), advertisements and community announcements in local media (e.g., television, radio, newspapers). The Cancer Council Victoria's Human Research Ethics Committee approved the study protocol. Subjects gave written consent to participate and for the investigators to obtain access to their medical records.
Men were excluded from the analysis if they had a confirmed diagnosis of prostate cancer prior to baseline (n = 106), had missing data on alcohol or covariates (n = 34), or were aged more than 70 years at baseline (n = 37). These exclusions left 16,872 men available for the analysis.
Assessment of alcohol and potential confounders
At baseline, a structured interview schedule was used to obtain information on potential risk factors including age, country of birth, education, smoking habits, and previous medical conditions.
To estimate alcohol consumption, participants were first asked if they had ever drunk at least 12 alcoholic drinks in a year; those that had not were considered lifetime abstainers. Nonlifetime abstainers were asked about their current average quantity and frequency of intake of beer, wine and spirits. Then they were asked about the intake of alcoholic beverages on each day during the week before the interview (diary). Subjects who were not lifetime abstainers, but did not consume alcohol at baseline were classified as exdrinkers. Total alcohol intake was categorized into lifetime abstainers, exdrinkers, 1–19 g/day, 20–39 g/day, 40–59 g/day ≥60 g/day following the guidelines of the Australian National Health and Medical Research Council (NHMRC).5 Wine, beer and spirits were categorized similarly, but exdrinkers were excluded from beverage-specific analyses and the highest 2 categories of wine and beer were combined and spirits were analysed as none versus any.
Cohort follow-up and ascertainment of prostate cancer cases and deaths
Cases were men notified to the Victorian Cancer Registry with a first diagnosis of adenocarcinoma of the prostate during follow-up to December 31, 2003. Gleason score or tumour grade were ascertained and used to categorise prostate cancer grade into low (Gleason score 2–4 or well-differentiated, N = 95, 13%), moderate (Gleason score 5–7 or moderately differentiated, N = 496, 68%) and high grade (Gleason score 8–10, poorly differentiated or undifferentiated, N = 123, 17%). Information on Gleason score or grade was not available in 18 cases (2%). Most cases had T1 or T2 tumours (N = 640, 87%), while only 13 cases (2%) had locally advanced or distant tumours (T4, N+ or M+). Information on stage was not available in 30 cases (4%).
High-grade tumours, locally advanced or distant tumours were grouped together as “aggressive” prostate cancer. Eleven prostate cancer cases had insufficient clinical information to determine tumour aggressiveness. Deaths and cause of death were identified from Victorian death records and from the National Death Index for those who died outside Victoria. Residential addresses were determined by record linkage to Electoral Rolls, from electronic phone books and from responses to mailed questionnaires and newsletters. By the end of the follow-up on December 31, 2003, 300 (1.8%) of the men included in this analysis were known to have left Victoria and were considered lost to follow-up for the incidence analysis; 24 (0.1%) were known to have left Australia and were considered lost to follow-up for the mortality analysis.
Follow-up began at baseline and continued until diagnosis of prostate cancer (death from prostate cancer for the mortality analysis), death, date left Victoria (Australia for the mortality analysis), or December 31, 2003, whichever came first. Cox's proportional hazards regression with age as the time axis was used to derive adjusted hazard ratios for men in each category of exposure. The model was adjusted for left censoring at the age at baseline.
Cox regression models based on competing risks were fitted using a data duplication method to test the heterogeneity in the rate ratios between aggressive and nonaggressive cancer.6 These analyses were performed stratifying the Cox models for type of failure (aggressive or nonaggressive prostate cancer) thus allowing the hazard function associated with the 2 types of failure to be different.
All final analyses included country of birth as covariate. Further adjustment for other potential confounders including education, body mass index, smoking, total energy intake, and previous medical conditions did not materially change the estimated rate ratios.
For the analyses of total amount of alcohol and type of beverage, we used data from the questions on frequency and quantity. The associations with type of beverage were evaluated by including consumption of wine, beer and spirits in the same model.
Data from the diary (intake of alcoholic beverages on each day during the week before the interview) were used to examine associations with drinking pattern. First, we analysed the number of drinking days during the week grouped as 1–3 days, 4–6 days and 7 days, separately for 2 levels of total consumption for the week (<140 g and ≥140 g). Second, we applied the multiple fractional polynomials algorithm to the number of drinking days and total alcohol consumption, forcing both variables into the model.7 Fractional polynomials are a method of analysing dose-response curves that makes no a priori hypothesis about their shape.8 One hundred and forty four subjects with missing data on alcohol consumption during the week before interview were excluded from the analysis of drinking pattern. Subjects with no intake of alcohol during the week before baseline were excluded from the analysis with fractional polynomials.
A sensitivity analysis was performed to investigate the effect of preattendance illnesses on the association between alcohol and prostate cancer in two ways: (1) excluding the first two yeas of follow-up; (2) including a time-dependent covariate in the Cox regression model that tests if the association between alcohol and prostate cancer differs with time since baseline.9
Statistical analyses were performed using SPLUS 6.2 (Insightful Corporation, Seattle, WA) and Stata 8.2 (Stata Corporation, College Station, TX).
Prostate cancer incidence
Among the 16,872 men eligible for the study, we identified 732 prostate cancers, including 132 aggressive tumours, during an average of 10.0 years of follow-up per subject and 53 prostate cancer deaths during an average of 10.3 years of follow-up. One hundred and ten (15%) prostate cancer cases were diagnosed during the first 2 years of follow-up. Table I shows the distribution of age, country of birth, and alcohol consumption by disease status and tumour aggressiveness.
Table I. Baseline Characteristics of Men in the Melbourne Collaborative Cohort Study by Outcome Status
|Age at baseline (years)|
| <50||35 (5.9)2||3 (2.3)||5135 (31.8)|
| 50–59||176 (29.9)||24 (18.2)||5069 (31.4)|
| 60+||378 (64.2)||105 (79.5)||5936 (36.8)|
|Country of birth|
| Australia/New Zealand||426 (72.3)||103 (78)||10503 (65.1)|
| United Kingdom||48 (8.1)||6 (4.5)||1330 (8.2)|
| Italy||66 (11.2)||19 (14.4)||2309 (14.3)|
| Greece||49 (8.3)||4 (3)||1998 (12.4)|
| Primary||98 (16.6)||17 (12.9)||3017 (18.7)|
| Started high/technical||207 (35.1)||46 (34.9)||5001 (31.0)|
| Completed high/technical||150 (25.5)||42 (31.8)||4004 (24.8)|
| Degree/diploma||134 (22.8)||27 (20.5)||4118 (25.5)|
| Never||244 (41.4)||47 (35.6)||6605 (40.9)|
| Current||67 (11.4)||9 (6.8)||2379 (14.7)|
| Ex||278 (47.2)||76 (57.6)||7155 (44.3)|
| ≤25.0||163 (27.7)||34 (25.8)||4483 (27.8)|
| 25.1–28.0||216 (36.7)||43 (32.6)||5697 (35.3)|
| >28.0||210 (35.7)||55 (41.7)||5949 (36.9)|
|Alcohol intake (g/day)|
| Lifetime abstainer||84 (14.3)||28 (21.2)||2221 (13.8)|
| Exdrinker||38 (6.5)||7 (5.3)||792 (4.9)|
| 1–19||262 (44.5)||51 (38.6)||7259 (45)|
| 20–39||114 (19.4)||25 (18.9)||3293 (20.4)|
| 40–59||51 (8.7)||11 (8.3)||1427 (8.8)|
| 60+||40 (6.8)||10 (7.6)||1148 (7.1)|
| No intake||231 (39.2)||62 (47.0)||5807 (36.0)|
| 1–19||260 (44.1)||48 (36.4)||7497 (46.5)|
| 20–39||66 (11.2)||11 (8.3)||1857 (11.5)|
| 40+||32 (5.4)||11 (8.3)||979 (6.1)|
| No intake||197 (33.5)||53 (40.2)||5111 (31.7)|
| 1–19||309 (52.5)||62 (47.0)||8901 (55.2)|
| 20–39||49 (8.3)||10 (7.6)||1304 (8.1)|
| 40+||34 (5.8)||7 (5.3)||824 (5.1)|
| No intake||366 (62.1)||87 (65.9)||9813 (60.8)|
| 1+||223 (37.9)||45 (34.1)||6327 (39.2)|
Prostate cancer incidence
Overall, alcohol was not associated with prostate cancer incidence; all rate ratios were close to unity. Alcohol consumers had rates of aggressive prostate cancer about 30% lower than lifetime abstainers, while all rate ratios for nonaggressive prostate cancers were close to unity but the difference in the hazard ratios for aggressive and nonaggressive cancers was not statistically significant (Table II).
Table II. Hazard Ratios for the Association Between Alcohol Consumption at Baseline and Incidence of Prostate Cancer1
|Total alcohol (g/day)|
| Lifetime abstainer||Reference||0.62||Reference||Reference||0.58|
| Exdrinker||1.15 (0.81,1.63)6|| ||1.32 (0.90,1.93)||0.73 (0.32,1.67)|| |
| 1–19||0.98 (0.79,1.22)|| ||1.09 (0.85,1.40)||0.67 (0.43,1.06)|| |
| 20–39||0.96 (0.75,1.23)|| ||1.04 (0.79,1.39)||0.72 (0.42,1.23)|| |
| 40–59||0.95 (0.69,1.29)|| ||1.02 (0.72,1.45)||0.67 (0.33,1.33)|| |
| 60+||0.94 (0.67,1.30)|| ||0.97 (0.66,1.42)||0.75 (0.36,1.54)|| |
|Alcohol by type of beverage7 (g/day)|
| No intake||Reference||0.84||Reference||Reference||0.26|
| 1–19||0.97 (0.81,1.17)|| ||1.02 (0.83,1.26)||0.76 (0.51,1.14)|| |
| 20–39||0.95 (0.72,1.25)|| ||1.02 (0.75,1.38)||0.65 (0.33,1.28)|| |
| 40+||0.98 (0.70,1.37)|| ||0.91 (0.62,1.35)||1.18 (0.59,2.35)|| |
| No intake||Reference||0.70||Reference||Reference||0.69|
| 1–19||1.02 (0.85,1.22)|| ||1.06 (0.86,1.29)||0.84 (0.57,1.25)|| |
| 20–39||0.97 (0.72,1.3)|| ||1.04 (0.75,1.43)||0.77 (0.39,1.54)|| |
| 40+||0.96 (0.69,1.34)|| ||1.00 (0.69,1.45)||0.69 (0.31,1.53)|| |
| No intake||Reference||0.13||Reference||Reference||0.98|
| 1+||1.14 (0.96,1.35)|| ||1.14 (0.94,1.37)||1.13 (0.74,1.72)|| |
As for total alcohol, all rate ratios for specific beverages were close to unity, and there was no persuasive evidence of different associations for aggressive and nonaggressive cancers (Table II).
The total alcohol intake during the week before the interview was highly correlated with the average daily intake from the quantity-frequency questions (Spearman's correlation, 0.85). During the week before baseline, 5008 (30%) men did not drink and were excluded from the analysis of pattern with fractional polynomials. Table III shows the hazard ratios in relation to the number of drinking days within categories of total weekly alcohol intake estimated from the diary. Compared with regular drinkers, the hazard ratios were 1.56 for those who consumed at least 140 g of alcohol in 1–3 days and 1.20 for those who consumed the same amount in 4–6 days. The best model selected with the multiple fractional polynomial algorithm included both variables as linear and showed almost no association with either the number of drinking days (hazard ratio per additional drinking day 0.97; 95% confidence interval (CI) 0.94, 1.02; test for trend, p = 0.21) or the average amount of alcohol consumed (hazard ratio per 10 additional g/day 1.01; 95% CI 0.97, 1.06; test for trend, p = 0.60). The interaction between alcohol consumed and the number of drinking days was not significant (p = 0.11).
Table III. Hazard Ratios for the Associations Between Pattern of Alcohol Consumption During the Week before Baseline Interview and the Incidence of Prostate Cancer1
|<140||1.04 (0.74, 1.44)4||0.93 (0.63, 1.38)||Reference|
|≥140||1.56 (0.91, 2.68)||1.20 (0.90, 1.60)||Reference|
Prostate cancer mortality
Fifty three prostate cancer deaths occurred during follow-up, including one during the first 2 years. No prostate cancer deaths were observed among exdrinkers, which were then excluded from the analysis of mortality. The associations between alcohol consumption and prostate cancer mortality are reported in Table IV. The minimum hazard ratio was 0.56 in the category 1–19 g/day and the maximum was 0.73 in the highest category of consumption, but overall alcohol was not significantly associated with prostate cancer death.
Table IV. Hazard Ratios for the Associations Between Alcohol Consumption and Prostate Cancer Mortality1
|Total alcohol (g/day)|
| Lifetime abstainers||Reference||0.916|
| 1–19||0.56 (0.28,1.14)2|| |
| 20–39||0.60 (0.26,1.38)|| |
| 40+||0.73 (0.32,1.70)|| |
Effects of undetected prostate cancers at baseline
The exclusion of the first 2 years of follow-up (men possibly with undetected prostate cancer at baseline) did not change the results materially for incidence or mortality and no statistically significant difference in the hazard ratios was observed by duration of follow-up (results not shown).
Overall, alcohol intake was not associated with prostate incidence or mortality. We did observe a slightly reduced incidence of aggressive disease and lower mortality from prostate cancer in men consuming alcoholic beverages compared to abstainers but the differences were not statistically significant and there was no evidence of a dose-response effect. Compared to men who drank on 7 days of the week before interview, men who consumed a high amount of alcohol during 1–3 days of the week had a 50% higher rate of prostate cancer, although overall no statistically significant association was found between pattern of alcohol and prostate cancer incidence.
Strengths of this study are its prospective design, low loss to follow-up (less than 2% of men left Victoria during the follow-up period) and close to complete ascertainment of cases (cases having been ascertained through the Victorian Cancer Registry). Also, because the information on alcohol consumption was collected at baseline, the possibility of bias due to differential recall is slight. Another strength of the study is that we have obtained information on tumour characteristics (i.e., stage or grade) for almost all cases. This is particularly important in Australia where, though no organised mass screening is recommended, PSA testing is regularly offered to men aged 50 years or older as happens in most developed countries.
Limitations common to all the studies that looked at the possible association between alcohol consumption and prostate cancer include the small number of cases of aggressive and fatal cancers and the lack of information about changes in drinking habits during follow-up, which could be important if the exposures have a late effect on prostate carcinogenesis. Because we did not collect information about the type of wine consumed, we were unable to test the hypothesis proposed by Schoonen et al. of a protective effect of red wine due to its high polyphenol content.3
Alcohol intake may have both adverse and protective effects on prostate carcinogenesis. The increased oxidative stress associated with high alcohol consumption might act early in prostate carcinogenesis, resulting in higher prostate cancer rates for drinkers.2 Our results are not consistent with this hypothesis but we cannot exclude an effect of very high consumption of alcohol that was not possible to test in our study because only a small percentage of men in the cohort consumed alcohol more than 60 g/day. Experimental evidence also suggests that ethanol alters the profile of sex hormones in a way that would reduce the growth of prostate cancer.2 The hormonal hypothesis implies that alcohol would have a protective effect by acting late in prostate carcinogenesis. The reduced incidence of aggressive cancer and mortality for those who drank alcoholic beverages observed in our study, albeit nonsignificant, is consistent with this hypothesis.
Men who consumed a total of 140 g or more of alcohol in 1–3 days of the week before the baseline interview had higher incidence than those who drank a similar amount during the week, but who drank every day. This finding, although not statistically significant, is similar to that previously reported by Platz et al.2 However, no significant association with pattern was seen in the analysis when the number of drinking days and the total amount of alcohol were treated as continuous variables. Further investigations are required to examine the effect of drinking pattern, especially at high doses of total alcohol intake.
In summary, we have found weak evidence that regular moderate alcohol consumption may protect against aggressive and fatal prostate cancer. In the era of PSA testing, when the majority of prostate cancers are of small volume and nonaggressive, very large studies are needed to be able to confirm a possible protective effect of regular moderate alcohol consumption on aggressive and fatal prostate cancer. If such an effect is confirmed, it will have a major impact on the prevention of prostate cancer because there are no established modifiable risk factors for this common type of cancer.
This study was made possible by the contribution of many people, including the original investigators and the diligent team who recruited the participants and who continue working on follow up. We would also like to express our gratitude to the many thousands of Melbourne residents who continue to participate in the study.
Cohort recruitment was funded by VicHealth and the Cancer Council, Victoria. The study was further supported by infrastructure provided by the Cancer Council, Victoria.