Beer and cancer: a risk worth worrying about?

Just after Christmas, on 27th December to be exact, several media outlets reported some research indicating that even a pint of beer a day increases the risk of bowel and liver cancers by a fifth. The person quoted was Dr Rachel Thompson, of the World Cancer Research Fund (WCRF), according to whom two units of alcohol a day increases the risk of bowel cancer by 18% and the risk of liver cancer by 20%.

These statistics sound quite frightening on the face of it, but the problem is that they do not give any indication of the absolute numbers involved. For instance, when we are told that a particular behaviour doubles our risk of death then we tend to get rather worried, but a doubling of risk might actually mean that the number of people actually dying increases from 1 in a million to 2 in a million. When put this way, it hardly sounds like anything to worry about.

Risk stories of the "beer causes cancer" kind usually emanate from press releases, so I went to the website for the WCRF and located the page listing their most recent press releases. I was hoping that some more detailed statistics would be found in the original press release, but as of 28th December the relevant press release was not on the website.

Some of the news stories did report that 3,000 people are diagnosed with liver cancer each year and 36,500 people with bowel cancer, figures that presumably also came from Rachel Thompson. Unfortunately, they did not indicate how many of these cancers were themselves caused by drinking. Nor did they indicate how many people in the UK consume alcohol, nor any indication of how the risks might accumulate over time (hence relate to age). This makes it practically impossible to properly assess the increase in cancer risk due to drinking on the basis of the information reported.

The news stories also did not say how the WCRF had come up with its figures. Was this on the basis of experimental testing with animals, in which case how well do animal results reflect the development of illness in humans? Or were the statistics based on comparing the incidence in cancer among drinkers and non-drinkers? And given that there are few drinkers (I would guess) who drink exactly 2 units of alcohol a day, how did they figure out that drinking exactly 2 units per day increases the risk of liver and bowel cancers by a fifth? My bet is that this statistic was inferred from some other statistic. But even more problematic is the likelihood that drinkers and non-drinkers differ in many ways apart from their willingness to imbibe alcohol. Thus, these other factors may be the causal factor in any increased risk for cancer. Of course, this may have been taken into account in the research, but in the absence of any description of the research how can we know?

Most disappointingly, the reporting of this story in the print edition of the Guardian came just one page after Ben Goldacre's Bad Science column, where he referred to the continuing use of relative risk figures in news stories.