Although alcohol is not known to be carcinogenic in animal experiments, there is strong epidemiological evidence that a high alcohol consumption increases the risk of cancers of the oral cavity, pharynx, larynx and esophagus. Although, less elevated, this also applies to the risk of stomach cancer, pancreatic cancer, colorectal cancer, cancers of the liver, lungs, breasts and the ovaries.1, 2, 3
It is still debated how the effect of different types of beverages diverges on the development of cancer. It has been hypothesised that beer consumption may increase the risk of rectum cancer, which could be explained by the nitrosamine contamination of beer, since nitrosamine is a well-known carcinogen,4, 5, 6, 7, 8 although an increased risk has also been related to spirits8 and wine.9
Since the foundation of the Danish Brewery Workers Union in Denmark in 1898 and until 2001, there has existed a formally recognised agreement between the Workers Union and the employers implying that the breweries should provide 6 bottles of beer to each brewery worker every day. A study from 1979 examined the incidence of cancer in a cohort of Danish brewery workers employed at least 6 months in the period 1939–1963 and observed that the cancer morbidity 1943–1972 was increased for cancer of the pharynx, esophagus, larynx and liver, while no increased risk for either cancer of the colon and rectum was observed.10
In our study, we investigated if members of this old and unique cohort with a high, prolonged daily consumption of beer are of increased risk of cancer in particular colorectal cancer. Compared to the previous study, we took advantage of 27 more years of follow-up, thereby shedding light of the association between high beer intake and development of cancers with a prolonged latency time.
Material and methods
Establishment of the cohort
The cohort consists of all male members of the Danish Brewery Workers Union (BWU) employed at least 6 months in a Danish brewery between 1939 and 1963. The BWU consists of the total work force in Danish breweries, employing 98–99% of all labour in the production and distribution of beer and mineral water in Denmark.
Of the 14,313 brewery workers from the original cohort, we were in possession of information on date of birth, the 3 first letters of the first name, the first letter of up to 2 middle names and the first 6 letters of the last name. In 1968, the Civil Register System (CRS) in Denmark was established, which implies that all live-born infants and all new residents in Denmark are assigned a Personal Identification Number (PIN). This number encodes sex and date of birth.
We identified the PIN of each brewery worker by extracting every person in Denmark with the same date of birth and first 6 letters of the last name as an identification of potential cohort members. In CRS, 42,223 persons were identical to this information. Two independent coders examined the extracted information to identify, if possible, the PIN of each brewery worker by considering the further information on the first name and 2 middle names. By this procedure a total of 10,468 of the original cohort of brewery workers were identified. Four of these brewery workers were registered twice in the original file because they had been employed at 2 different breweries. Only the first membership period was included in our study.
On 2,588 of the remaining 3,845 brewery workers without an identified PIN, we were in possession of information about date of death from the original file.10 All 2,588 brewery workers died between 1943 and 1974. We matched those brewery workers against the total Cancer Registry file for 1943–1974; for each brewery worker, all registered male cancer cases with identical day of birth and death were extracted. This procedure resulted in 810 matches. These potential matches were then scrutinised visually for consistency between first, middle and last names, leaving 694 primary cancer diagnoses. One brewery worker was filed twice in the original file but was only included once in the study base.
The entire study population therefore comprised 13,051 brewery workers (91.2%) (Table I).
Table I. Cohort of Male Members of the Danish Brewery Workers' Union Followed-Up from 1943–1999 (Percentage in Parenthesis)
Beer production workers
Mineral-water factory workers
Brewery workers with PIN identified
Brewery workers with date of death
Reasons for exclusion
Beer consumption by Danish brewery workers and the general Danish population
Since the foundation of the BWU in 1898, an old tradition of providing free beer to brewery workers by the breweries was formally recognised in agreements between the BWU and the Brewers Association (the employers organisation). The agreed ration, unchanged since the original agreement and until 2001, was set at 6 bottles of beer for each brewery worker per day, i.e., each brewery worker received a free ration of 2.1 litres of beer per day. This amount of beer provided to brewery workers had to be consumed on the premises of the brewery and, for tax reasons, could not be brought out of the breweries.
The free offered beer was of the light pilsner (lager) type containing 3.7 g ethanol per 100 ml. An average brewery worker therefore had the possibility of consuming 77.7 g of ethanol per day at work. Both the BWU and the Brewers Association confirm that this average was consumed.10
On the basis of statistics on alcohol consumption in Denmark,11 the average daily consumption of beer per person among adult Danes were: 107 ml in 1930; 89 ml in 1940; 147 ml in 1950 and 163 ml in 1960. These figures are, however, not directly comparable with the average daily consumption of 2,100 ml by brewery workers, as the statistics give no information on sex differences in beer consumption and possible differences in levels of consumption according to place of residence.
For purposes of comparison with male brewery workers, it is important to estimate the beer consumption of the general male population. In a survey of Copenhagen males aged 30–45 in 1964, the average daily alcohol consumption of the population sample was 1.33 units/day composed of beer, wine and spirits. It can be estimated that average beer consumption in the survey sample was between 310 and 410 ml of beer per day.12 In another dietary survey of males aged 55–64, the average daily consumption of beer in a random sample of Copenhagen males was found to be 450 ml of beer per day, i.e., 1.3 bottles per day.13
Because the majority of the members of BWU are employed and live in Copenhagen, it may be concluded that Danish brewery workers have a high daily consumption of beer, estimated to be at least 4 times higher than that for the average Danish males.10
Among the members of the BWU, only 10,901 of the workers (76.2%) were employed in beer brewing industries (beer production workers) and therefore had the right to free beer. On the other hand, 1,063 workers (7.4%) were employed in factories producing mineral waters (mineral-water factory workers) and had no free ration at their disposal. The latter may therefore be assumed to consume alcohol, including beer, at the same level as the general population of the socio-economic stratum to which members of the BWU belong. We were able to identify 9,963 and 953 of these 2 groups, respectively (Table I). The remaining of the original cohort held employments both in the beer brewing industry and in a mineral water factory, or worked as dray-men, representing a selected subgroup of brewery workers; the drinking patterns of this subgroup is ill defined, and they have therefore not been included separately in the analyses (see Table III).
Table III. Incident Malignant Neoplasms of the Buccal Cavity and Pharynx, Larynx, Esophagus, Colorectum, Pancreas, Urinary System and Lungs Among Male Beer Production and Mineral Factory Workers 1943–1999 Compared to the General Male Population of Denmark
Beer production workers
Mineral-water factory workers
Buccal cavity and pharynx
Lung primary, trachea
We identified the subjects who developed cancer during follow-up through record linkage between the 10,464 brewery workers with identified PIN and the Cancer Register. We excluded subjects with a history of cancer before the unionisation date (n = 12) and subjects with missing values on date of unionisation (n = 6). For brewery workers with more than 1 cancer diagnosis, only the first diagnosis was included in our study. Site-specific cases of cancer in the cohort were classified with the original 4-digit ICD-7 code by which the case appears in the Cancer Registry. These codes was collapsed into broader categories according to separate organs and organ systems.14
For brewery workers with identified PIN, vital status was followed until 31 December 1999 in the Civil Registration System (n=10,464), while for the 2,587 brewery workers without identified PIN, the date of death were known from the original file. The observation period for each subject was from 1 January 1943 when the Danish Cancer Registry was established or from date of unionisation, whichever was the latest until time of death, emigration, or disappearance, time of diagnosis of cancer or 31 January 1999, whichever came first; 0.5% of our study sample (n= 63) disappeared or emigrated. The total person-years of observation for the brewery workers' cohort were 422,240 person-years.
Registration in the Danish Cancer Registry is based on notification forms that are completed by hospital departments (including departments of pathology and forensic medicine) and practising physicians whenever a case of cancer is diagnosed or found at autopsy and whenever there are changes in an initial diagnosis. Cases recorded manually are supplemented by unrecorded cases revealed by computerised linkages to the death certificate files and the National Patient Register.15 The entire process is supervised by medical doctors. Ambiguous or contradictory information, either within a notification form or between forms, leads to queries in approximately 10% of notifications received. Comprehensive evaluation has shown that the Register is 95–99% complete and valid.16, 17, 18, 19
Expected numbers of cancer cases and calculation of standardised incidence ratios
We derived expected numbers of cancer cases by multiplying age- (5-year groups) and sex- (males) specific incident cases and corresponding man-years at risk in intervals of the follow-up period taking account of possible time trends in disease occurrence. By summarising over the various cells this results in the expected numbers of cancer cases indirectly standardising for age and time trends.
The standardised incidence ratios (O/E) were estimated as the ratio between observed and expected numbers of cancer cases. Tests of significance and 95% confidence intervals were derived assuming that the observed numbers followed a Poisson distribution and calculated by formulas given by Breslow & Day.20
Subsequently, the 9,963 workers employed in beer brewing industries with the right to free beer and the 953 workers employed in factories producing mineral waters only with no right to free beer were analyzed separately. Incidence of the alcohol associated cancers of the buccal cavity and pharynx, larynx and esophagus, colorectum, pancreas, urinary system and lungs for these 2 groups were compared to the expected number of cancers.
During the follow-up period 1943–1999, a total of 3,928 cases of cancer was observed in the brewery workers cohort compared to 2,835.8 cases expected yielding a O/E of 1.39 (95% CI, 1.34–1.43) (Table II). Strongly excess risks of developing cancer are noted for cancer of the tongue (O/E=2.63), cancer of the mouth (O/E=2.43), cancer of the pharynx (O/E=3.09), cancer of the esophagus (O/E=3.06), cancer of the liver (O/E=3.02) and cancer of the larynx (O/E=2.68), while more moderately elevated risks are seen for cancer of the gallbladder (O/E=1.66), cancer of the pancreas (O/E=1.34), cancer of the lungs (O/E=1.67) and cancer of the urinary system (O/E=1.21). A barely insignificantly decreased risk is noted for the development of skin cancers other than melanoma (O/E=0.90).
Table II. Incident Malignant Neoplasm 1943–1999 Among Male Members of the Danish Brewery Workers' Union Compared to the General Male Population of Denmark
Site (ICD 7)
All cancers (140–205)
All cancers minus other skin
Buccal cavity and pharynx (140–148)
Salivary glands (142)
Digestive organs and peritoneum (150–159)
Small intestine (152)
Colon incl. recto sigmoid (153)
Rectum excl. anus (154)
Gallbladder, bil. pass., amp. vater (155.1)
Respiratory system (160–164)
Lung primary, trachea (162.0,1)
Male genital organs (177–179)
Urinary system (180–181)
Melanoma of skin (190)
Other skin (191)
Other specified sites (192–197)
Lymphatic and haematopoietic tissue (200–205)
Moderately but significant elevated risks of cancer of the colon (O/E=1.26) and cancer of the rectum (O/E=1.31) are also observed.
Table III shows the standardised morbidity ratios of 7 groups of alcohol-associated cancers, contrasting the experience of workers employed in beer production and in the mineral-water factories, which indicates that there are differences between brewery workers with a high daily beer consumption and workers without this privilege for several of these cancers. For beer production brewery workers the O/E is significantly elevated for all 7 groups, most strongly for esophagus, larynx, buccal cavity and pharynx and lung cancer. The elevated risk of esophageal and colorectal cancer is totally absent for mineral-water factory workers, while the risk is insignificantly elevated for cancer of the buccal cavity and pharynx, larynx and pancreatic cancer. In general, the insignificant results among mineral-water factory workers may be due to few cancer cases.
For cancers of the urinary system, beer production workers have a slightly lower point estimate than mineral-water factory workers, which might be explained by chance. For lung cancer the risk estimates are almost identical for the 2 groups of brewery workers.
Our study shows that Danish brewery workers have an elevated risk of developing cancer, especially cancers of the buccal cavity, the digestive tract and the respiratory system. The incidence rate of cancers of the tongue, mouth and pharynx, esophagus, liver and larynx is strongly elevated, while the risk of pancreas cancer, cancer of the colon and rectum, cancer of the lungs and cancers of the urinary system are weakly elevated. The difference in alcohol-related cancer morbidity between workers employed in beer production and in the mineral water factories, respectively, supports that high beer consumption may be a causal component for the development of cancer.
These results are generally in congruence with previous studies among brewery workers.10, 21, 22 The previous report on this cohort10 observed weaker risk estimates compared to our study, e.g., cancer of the esophagus in the previous study was significantly elevated (O/E=2.09; 1.50–2.83), while the estimate in our study is even stronger (O/E=3.06; 2.54–3.66). This pattern is also observed for cancer of the buccal cavity and pharynx, the liver, colon and rectum cancer, cancer of the larynx, lung and urinary system.
Figure 1 illustrates that 3 of the studies on brewery workers observed significantly increased risks of rectal cancer, while the first Danish study revealed no increased risk.10 The risk pattern for colon cancer is very similar across the 4 studies, although the results from our study is the only one being statistical significant.
The differences between the 3 previous investigations for cancer of the colon and rectum have been ascribed to different contents of the beers4 and to the considerable higher incidence rate of rectal cancer in the general Danish population compared to the Irish and the Swedish populations. For this reason, it has been hypothesized that the absolute risk for rectal cancer would be more homogeneous among the studies.22 On the other hand, the results of our study give strength to another explanation. The latency time from the exposure of a carcinogenic substance to the clinical pronounced colorectal cancer may be decades, which is conceptualised by the adenoma-adenocarcinoma sequence that is widely accepted as the primary development of colorectal cancer.23, 24 Therefore it is possible that the excess number of observed cases seen in our study was not developed to a clinical level at the time of follow up in the first Danish study (1972). To further evaluate this explanation, we studied whether the O/E-ratio changed as the latency time increased. This was done by including risk-free time-windows of 20, 25, 30, 35, 40, 45 and 50 years after unionisation date for brewery workers with unionisation date after 1 January 1943 (n=8,398). The O/E-ratios of rectum cancer were slightly higher for longer latency times, with maximum O/E-ratios between 30 to 45 years after unionisation date (data not shown). This further strengthens the latency time explanation.
It is recognised that consumption of alcohol and smoking habits are strongly correlated and therefore it is possible that our findings are confounded by the carcinogenic effect of smoking. Especially cancers of the upper aerodigestive organs are related to tobacco as well as to alcohol. The effect of smoking could not be investigated in our study because no information on smoking habits was available for the cohort. Tobacco smoking however was prohibited on the premises where beer (and mineral water) was produced.10
The elevated O/E-ratios for lung cancer among both beer production and mineral-water factory workers suggest that smoking may be a confounder. The weak associations found for cancers of pancreas and urinary system may be explained by confounding by smoking. Methodological studies have concluded that only substantially different distributions of the confounding factor among the exposed group would cause a considerable confounding effect,25, 26 and furthermore a number of studies have shown that alcohol consumption after adjustment for tobacco smoking is a risk factor of lung cancer.27, 28 This indicates that tobacco is not likely to be a severe confounder of the associations found in our study.
It is not possible in our study to preclude the possibility that other risk factors for cancer have influenced the findings. Besides smoking factors, such factors as dietary habits and social status, could also have influenced the associations. We have no information about the dietary habits of the cohort, while concerning social factors, we know that the cohort is very homogeneous because the members of the Brewery Workers Union presumably have almost the same income and education. But in comparison with the general male population of Denmark, this internal homogeneity may potentially influence the external comparisons carried through in our study. Brewery workers could have a more harmful pattern of exposures to other risk factors and our results could therefore be confounded. But it could be reasoned that only very healthy workers are able to live through a long working life with a high daily consumption of beer (Healthy workers' effect). It might therefore be presumed that our results are biased towards zero. The elevated risk of alcohol associated cancers might have been even stronger if we could have adjusted for potential confounders.
The measurement of consumption of beer in the cohort is not individually recorded but instead our analyses rely on an ecological measurement. We know that each worker had the opportunity to consume 6 beers each day and that they presumably did so, but we have no information about the exact consumption of each brewery worker. If some of the brewery workers consumed less than 6 bottles of beer each day, we would expect that their risk of developing cancer would approach the risk of the general male population of Denmark. Therefore it would be expected that the standardised incidence ratios would be underestimated. In addition, our estimates are conservative in the sense that our results (Table II) are influenced by the inclusion of a group of workers employed in the mineral water factory in which beer consumption was not allowed on the premises of the brewery.
Through the establishment of the cohort it was not possible to identify the cancer morbidity of 1,257 brewery workers (8.8%), due to missing PIN or date of death. If these brewery workers for some reason had another risk of developing cancer than the identified brewery workers, the findings of our study could have been either positively or negatively influenced. To evaluate this potential selection bias, the information available for all brewery workers from the original cohort was consulted: the date of birth and date of joining the BWU. The median date of birth for the brewery workers included in our study was approximately 5 years earlier compared to workers not identified and the median date of joining the BWU was approximately 3 years earlier for the same group. These differences could indicate that the brewery workers identified are healthier because they presumably have lived longer than the brewery workers not identified. Therefore the effect found in our study might be an under estimate of the true effect.
In conclusion heavy beer drinkers are at high risk of cancers, especially for cancer sites known to be associated with a high consumption of alcohol.
We thank the Danish Cancer Society for availability of the original data files and Visti Birk Larsen for valuable comments on construction of the cohort.