Tuesday 30 December 2014

Communicating oral cancer risk due to alcohol

A little while ago I tweeted about a systematic review on alcohol and oral cancer risk that the dental elf had flagged up.
But how can be communicate risk in a balanced, non-alarmist, but hopfully behaviour changing way?

In keeping with my previous blog about shared decision-making, patients are entitled to make choices about how they live their lives. In order to do that they may want to have access to some numbers that help them to take a proportionate approach.

In the review of risk of oral cancer the alcohol intake was classified as such:
Light, moderate and heavy drinking was respectively ≤12.5, ≤50 and >50 g per day of alcohol
1 unit of alcohol is the equivalent of 8g, or 10ml, of pure alcohol. (There's a handy guide for those considering their alcohol intake - or wishing to help others to make a change - here. Picture of alcohol units is taken from this)


So roughly the:

  • light intake was up to 1.5 units per day
  • moderate was up to 6 units 
  • heavy was over 6 units per week

The Dental Elf drew attention to the different types of studies included in the review. When only cohort studies were included in the meta-analysis, which are at lower risk of bias, the risk of developing oral cancer was:
  • no greater for the light drinkers, 
  • about one quarter higher for the moderate drinkers and 
  • about 3 times greater for the heavy drinkers. 

The Elf's table reproduced below shows that if case-control studies alone are considered the risks appear much greater. As there is a higher risk of bias in these studies I am inclined to concentrate on the cohort studies alone.


As those of you who are familiar with the work of Gert Gigerenzer, among others, relative risks are not great at communicating risks in a way that allows a proportionate response from a patient yet a lone a clinician.

To say that there is a 3x greater risk of oral cancer if you drink more than 6 units of alcohol a day may sound a lot. But if the lifetime risk of getting oral cancer were, say 1 in 100,000 then having a risk of 3 in 100,000 may not seem terribly significant (assuming there were no other health or social risks to drinking that much).

What we really need to be able to better understand the implications of the increased risk is what the absolute risk of getting oral cancer is.

The closest I could come to this is an overall lifetime cancer risk calculated by the Statistical Information Team at Cancer Research UK. They reckon that 1 in 84 people will develop oral cancer. The number caused by smoking is estimated to be about a half and a third due to alcohol (stats from cancer research). So this number will not be a true reflection of the risk for a non-smoker and non-drinker, which would be expected to be much lower. Can we make use of it though?

My view is that it is not the detailed number that matters when communicating the risk but an indication to a patient of the scale of the impact of their behaviour. So I would recommend making an assumption based on the 1 in 84 figure.

Let's assume that 1/3rd of cancers are caused by alcohol as Cancer Research suggests and that 1/2 are caused by smoking. As a few more are caused by smoking let's assume the average of 1 in 84 is a little lower for drinkers - say 1 in 100. Let's then assume that as about 1/6 of the cancers are in non-drinkers and non-smokers that the lifetime risk in this group is about 1 in 200.

Now we can try to give a sense of the impact of drinking more than 6 units of alcohol by saying:
In non-drinkers we would expect approximately 1 in 200 to get oral cancer. For those who drink more than 6 units per day the number would be 3 in 200.
Now, as we would like to communicate the increased risk for the moderate drinkers of 25% we need to increase the size of the denominator by 4 so that we can say:
In non-drinkers we would expect approximately 4 in 800 to get oral cancer. For those who drink between 1.5 and 6 units a day we'd expect 5 in 800 to get oral cancer. For those who drink more than 6 units a day we'd expect 12 in 800 to get oral cancer.
I hope that this will help communicate oral cancer risk in a balanced way.

No comments:

Post a Comment