Addiction recently published a summary of the evidence linking alcohol consumption to cancer.  The author, Jennie Connor, was not presenting new data; however, just 13% of people mention cancer when asked about the risks of drinking, so Connor’s essay attracted huge public and media interest. In doing so, it raised questions about how we communicate the risks of drinking and, particularly, where the message ‘no safe level of drinking for cancer’ is taking us.

From a scientific perspective, the central point of Connor’s case is watertight.  Meta-analyses consistently demonstrate a relationship between how much people drink and their risk of developing seven cancer types: oropharynx, larynx, oesophagus, liver, colon, rectum and breast (in women).  In each case, the risk increases with each extra drink consumed and, although the risk is small at low levels of consumption, this has prompted the ‘no safe level’ message advanced by some public health advocates.  Although the biological pathways by which alcohol affects cancer development are not well understood, there are plausible options and, to my knowledge, no authoritative body is disputing the conclusion that alcohol can cause cancer.

Connor also points to scepticism regarding the beneficial effects of moderate drinking for cardiovascular health.  A steady stream of studies has shown that limitations in epidemiological methods may be making moderate drinking look healthier than it really is.  These limitations include:

  • Classifying former drinkers with poor health profiles as abstainers;
  • Comparing moderate drinkers to abstainers – a generally unhealthy bunch even after accounting for former drinkers;
  • Only including moderate drinkers who are alive and well when epidemiological data are collected;
  • Only measuring alcohol consumption at one time point so people with different drinking trajectories across their life course get mixed together;
  • Only measuring average daily consumption so people who do and do not binge drink to varying degrees get mixed together.

Throw in faintly implausible studies which suggest alcohol is protective against everything from liver cirrhosis to hearing loss and the failure to pin down a definitive biological pathway for the cardioprotective effect (although plausible candidates continue to be investigated) and you can see where sceptics regarding protective effects are coming from.

There are important caveats though.  While the benefits of drinking are likely to have been overestimated, there is insufficient evidence to state conclusively that they do not exist.  Moreover, the research limitations above affect all health outcomes, not just cardiovascular disease.  Other major biases, such as underestimation of consumption among drinkers, may also lead to overestimation of the risk associated with particular consumption levels.  Given these uncertainties, it is difficult to confidently state the risk or benefit of any level of drinking except in general terms.

So where does this leave communication of alcohol-related risks to the public?

The ‘no safe level’ message is increasingly prominent and does have a compelling clarity, but it raises at least four concerns.  First, there is a selectivity to the no safe level message as few alcohol-attributable health conditions (protective effects aside) have a no-risk threshold.   This suggests the focus on cancer may be less about unusually high risks at low consumption levels and, at least in part, more about the elevated position of cancer in the public mind and in policy processes.  Second the message fails to distinguish between the presence of risk and the scale of risk. Cancer Research UK say that drinking at low levels “won’t make much difference to an individual’s absolute risk of developing cancer” and their infographic below gives you a sense of the absolute risk for mouth cancer – a 0.5% risk if you don’t drink increases to a 0.6% risk if you drink a small glass of wine (125ml or about 1.5 units) a day.  The increase in risk for breast cancer is similar (11.1% to 11.7%) and remember this is for a glass of wine a day while the no safe level message is about any drinking.


Third, the uncertainty in epidemiological risk estimates means such small risks at low consumption levels may not be an appropriate focus for health promotion messages.  Fourth, and perhaps most worrisome, is where the no safe level message leads us.  I took part in a three-way radio interview about the Addiction paper with Jennie Connor and a spokesperson for the World Cancer Research Fund.  The latter stated (twice) that: “our message, for cancer prevention….is to not drink alcohol at all“.  This is a soft and caveated abstinence message but an abstinence message nonetheless.  It is one of the logical conclusions of the no safe level argument but not one, I suspect, that public health actors would unanimously support.

At the other end of the scale, I am wholly unpersuaded by the minority who call for moderate drinking to be promoted as part of a healthy lifestyle.  The evidence is too uncertain, the scale of benefit too unclear, the competing harms too many and too poorly understood by the general public, and the principal beneficiaries (i.e. middle-aged women) too select.  Moreover, people need little encouragement to drink and abstainers have a wide range of alternative options available for improving their cardiovascular health.

It is clearly difficult to construct health promotion messages that are scientifically robust, sufficiently compelling and easily understood.  We should also not forget that there are tensions in play between individual and population strategies to improve public health.  However, on-going debates around cancer and heart disease messaging and the drift towards abstinence-oriented advice suggests we should be alert to the risks epidemiology creates as well as those it uncovers.