Genetic evidence suggests that alcohol’s effect is limited to certain cancer types

Research published in BMC Medicine demonstrates a link between greater alcohol consumption and higher risk of some cancer types. However, the evidence was mixed, with the research indicating that alcohol consumption raises the risk of cancer of the head and neck, cancer of the oesophagus, and cancer of the bowels – parts of the body that alcohol comes into contact with when consumed. For other cancer types, evidence for a harmful effect was either inconsistent or absent.

Many claims have been made about the health risks of alcohol consumption from observational studies. However, such associations are subject to bias from various sources – does alcohol consumption increase or reduce the risk of disease, or is it that alcohol drinkers are healthier or less healthy than non-drinkers? This is a classic question of correlation versus causation. As it is difficult to conduct long-term randomized studies of alcohol consumption, it is important to investigate any potential link with cancer risk using a variety of approaches. These include so-called “natural experiments”, in which we consider factors leading to natural variability in alcohol consumption levels. The International Agency for Research on Cancer has highlighted seven cancer types which they claim have convincing evidence for a harmful effect of alcohol consumption: breast, bowel, oesophagus, liver, mouth, throat, and voice box. Based on this evidence, the United States Surgeon General recently recommended updating advisory warnings on alcohol sold in the US to inform drinkers of the link between alcohol consumption and cancer.

A team led by researchers at the University of Cambridge and Karolinska Institute studied data on over 1.4 million volunteers from the UK, Finland, and the United States using a technique known as “Mendelian randomization”. Participants were compared in a statistical analysis not according to their measured level of alcohol consumption, but according to whether they inherit genetic variants that on average predispose people to drink more or less alcohol. By making comparisons based on genetic variations rather than alcohol consumption levels directly, several of the pitfalls of judging causation from observational data are avoided, as the genetically-defined groups compared are similar with respect to factors other than alcohol consumption. For example, although alcohol consumption is correlated with household income, genetic predisposition to alcohol consumption is not. Hence statistical associations demonstrated by Mendelian randomization more plausibly reflect the impact of alcohol consumption.

A genetic score associated with an increase of around 3 additional drinks per week in alcohol consumption was linked to a 51% increase in risk of head and neck cancer (including cancers of the mouth, throat, and voice box), a 21% increase in risk of bowel cancer, and a 42% increase in risk of oesophagus cancer. These three cancer types comprise around 1 in 6 of new cancer cases in the UK per year. The genetic score was linked to a 40% increase in risk of liver cancer, but the statistical strength of evidence was only convincing in a dataset of American military veterans with high average levels of alcohol consumption. The genetic score was also linked to a 44% increase in the risk of cancer mortality.

However, there was no evidence indicating a harmful effect on risk of breast cancer either in the primary datasets with 51,000 breast cancer patients or in breast cancer-specific datasets with 133,000 additional patients. In addition, there was no consistent evidence suggesting a harmful effect of alcohol consumption on any other cancer type. Unexpectedly, the genetic score was linked with lower risk of kidney cancer, myeloma, and non-Hodgkin’s lymphoma. An inverse association between alcohol consumption and these cancers has also been observed in previous observational studies. While these associations suggest that alcohol consumption may lower the risk of some cancer types, these findings should be regarded with caution until they are confirmed using other study designs.

Overall, the investigation confirmed previous findings that alcohol increases risk of certain cancers, but it could not confirm an effect on breast cancer, and it suggested that the harmful effect of alcohol on cancer is limited to these cancer types. These cancer types are in parts of the body that come into direct contact with alcohol when consumed. Alcohol consumption does not appear to be a universal risk factor for all cancers.

Limitations of the investigation include uncertainty in the extent to which the genetic variants tell us specifically about the impact of alcohol consumption, the potential for alcohol consumption to affect cancer discovery rather than cancer risk, and the difficulty in distinguishing those who died “from cancer” and those who died “with cancer”. The investigation is unable to distinguish between the effects of light drinking versus heavy drinking, and the impact of different types of alcohol. It is also unable to test if alcohol consumption increases the severity of cancer.

Stephen Burgess, a Group Leader from the MRC Biostatistics Unit, who oversaw the study, said: 

"Our investigation confirms the harmful effect of alcohol consumption on cancer risk, but suggests that alcohol only increases cancer risk in body parts with direct exposure to alcohol. In particular, we did not see evidence supporting an effect of alcohol consumption on breast cancer risk, contradicting previous research."

Susanna Larsson, an epidemiologist based at the Karolinska Institute, who led the study analyses, said: 

"We could not confirm the link between alcohol consumption and breast cancer risk in our genetic study. This could mean the true effect is very small, or that previous findings were influenced by other factors, such as hormone use, which are difficult to fully account for in observational studies."

Siddhartha Kar, a physician-scientist based at the University of Cambridge, who contributed to the study, said:

"While drinking alcohol, particularly binge drinking, is broadly harmful to human health and society, increasing the risks of heart and liver diseases, dementia, suicide, and violence, the impact of alcohol on cancer is likely more limited."

Read full paper: Alcohol consumption and risk of cancer: a Mendelian randomization analysis of four biobanks and consortium data | BMC Medicine

More on Mendelian randomization: http://www.mendelianrandomization.com/index.php/what-is-mr