Alcohol consumption and cancer risk

This meta-analysis investigates the association between alcohol consumption and the risk of 18 neoplasms.

Bottom line

Alcohol consumption is associated with increased risk of some, but not all cancers, most notably cancers of the oral cavity and pharynx, oesophagus, larynx, breast and liver.

Reference

V Bagnardi et al. A meta-analysis of alcohol drinking and cancer risk. British Journal of Cancer 2001 85: 1700-1705.

Search

The literature was searched using several bibliographic databases (MEDLINE, Current Contents, Embase, CAB Abstracts and Core Biomedical Collection) from 1966 to 2000. References from identified papers, literature reviews and other meta-analyses were also examined and the most relevant epidemiological and medical journals were manually checked. Included were case-control or cohort studies published as original articles, reporting odds ratios or relative risks for at least three levels of alcohol consumption.

A total of 235 studies with 117,471 cases were identified. Of these, 187 were case-control and 48 were cohort. Eighteen cancer sites were investigated or all cancers irrespective of site. For each site, the number of studies varied from one to 49 and the number of cases ranged from 81 to 44,033. For melanoma and cancers of the small intestine, gallbladder, cervix and kidney, the risk estimates were based on only one or two studies. Three levels of alcohol consumption were examined: 25 g a day, 50 g a day and 100 g a day (25 g corresponds to approximately 2 drinks a day).

Results

Associations were found between each level of alcohol consumption and cancers of the oral cavity and pharynx, oesophagus, larynx, breast, liver, colo-rectum and stomach. Associations were also found between alcohol consumption and cancers of the ovary and prostate, but only for 50 g and 100 g a day. Table 1 shows the relative risks for each site.

No associations were found between alcohol consumption and cancers of the lung, small intestine, bladder, kidney, pancreas and endometrium. For melanoma and cancers of the gallbladder and cervix, no studies examined the effects of 100 g/day and no associations were observed at the other levels.

Table 1. Relative risks of alcohol consumption for each cancer site.

Cancer Site	Cases	Relative risks
		25 g day	50 g day	100 g day
Oral cavity and pharynx	7,954	1.8	2.9	6.0
Oesophagus	7,239	1.5	2.2	4.2
Larynx	3,759	1.4	1.9	4.0
Breast	44,033	1.3	1.7	2.7
Liver	2,294	1.2	1.4	1.9
Colon and rectum	11,296	1.1	1.2	1.4
Stomach	4,518	1.1	1.2	1.3
Ovary	1,651	no association	1.2	1.5
Prostate	4,094	no association	1.1	1.2

Whether associations were modified if results were adjusted for smoking was examined for cancers of the oral cavity, oesophagus, larynx, lung and bladder. Significant differences were found between smoking-adjusted and unadjusted results for cancers of the larynx, lung and bladder, but not for cancers of the oral cavity and oesophagus. For laryngeal cancer an alcohol-related risk persisted after adjustment for smoking (relative risks 1.29, 1.68 and 2.79) while for cancers of the lung and bladder it did not.

The effects of gender were examined for each site. Significant effects were only found for oesophageal and liver cancers, with higher risks for women.

Eight studies reported the relationship between alcohol and the risk of all sites together. Significant effects of alcohol started from 28 g a day.

Comment

While Table 1 provides a reference for the effects of alcohol consumption on the risk of cancers, several limitations mitigate a straightforward interpretation of these estimates.

Although the analysis included a total of 117,471 cases, for some cancers the number of cases was small. For melanoma and cancers of the small intestine, gallbladder, cervix and kidney case numbers ranged from 81 to 921. Furthermore, the number of cases at each level of consumption was not reported, but in some instances are likely to be small (e.g. high levels of alcohol intake for breast and ovary cancer in women).
Whether studies accounted for other lifestyle factors, e.g. diet, was not reported. Risk estimates should therefore be treated as a guide as other lifestyle factors may be responsible for some of the observed effects. Indeed, for the low risk estimates (i.e. around 1.1) it is possible that no associations would be found. (However, it is usual that alcohol consumption is assessed by self-report questionnaires, so a degree of under-reporting may have taken place which would result in an underestimation of the observed associations.)
The association between alcohol and liver cancer is difficult to interpret since most alcohol-related liver cancers follow cirrhosis which leads to reduced alcohol consumption.

Nevertheless, alcohol consumption increases the risk of cancers of the oral cavity and pharynx, oesophagus, larynx, breast and liver. It seems that moderate alcohol consumption (up to two drinks a day) will not appreciably increase the risk of cancers of the colo-rectum, stomach, ovary and prostate. The estimated risks associated with higher consumption levels for these cancers are more difficult to interpret. Compared with moderate drinkers, heavy drinkers are more likely to have unhealthy lifestyles; so estimates at the higher levels of intake are more likely to suffer from confounding by poor diet and other lifestyle habits than estimates at the moderate level. While this is true for all cancers, the increased risks associated with cancers of the colo-rectum, stomach, ovary and prostate are lower and therefore could become very low or not persist at all if other lifestyle factors were considered.