Oral cancer statistics

Today's news on the latest oral cancer statistics contains some schoolboy errors in presenting statistical results, and are a great example of how not to present statistics in the popular media.

Let's take the title of the article to start with:  "Drink blamed for oral cancer rise". Well, it's true that oral cancer is more common now than it was in previous decades. It's also true that we drink more now than in previous decades. And it's true that alcohol consumption is a risk factor for oral cancer. So it seems logical to assume that drink must be responsible, doesn't it?

Logical, maybe, but also wrong. This is a great example of the ecological fallacy. There are many other risk factors for oral cancer, some known, some no doubt remaining to be discovered. To assume that alcohol is the main cause of the rise in oral cancer is no more than a guess. Maybe it's the cause, maybe it's not. We just don't know. But of course journalists don't really like stories saying "we don't know", so sadly that's not how it's reported in the popular media.

As an aside, it's interesting to note the very strong interaction between tobacco and alcohol as risk factors for oral cancer. In layman's terms, if you smoke you're more likely to get oral cancer, if you drink you're more likely to get oral cancer, but if you do both then you are much more likely to get oral cancer, more so than you would think from simply combining the 2 separate increases in risk. Maybe one reason why oral cancer is becoming more common is that smoking and drinking are more likely to be associated with each other than they used to be. I have no idea whether they are or not, and would be interested to see any statistics on that, but if they are then that would certainly be expected to lead to an increase in oral cancer incidence.

The second thing in the news that I consider very poor style when reporting statistics is the focus on relative, rather than absolute risks. We are told "Numbers of cancers of the lip, mouth, tongue and throat in [men and women in their forties] have risen by 26% in the past decade" and we are further told that this is an "alarming" rise.

It sounds alarming, doesn't it? Put it in terms of absolute risks, however, and it seems a bit different.  The incidence of oral cancer in those in their forties rose from 7.1 per 100,000 in 1995 to 10 per 100,000 in 2006. Still find it alarming? I don't. That means there is still only a 1 in 10,000 chance that a person in their forties will develop oral cancer in any given year. I wonder how that compares with your chances of being run over by a bus?

Yes, it's not good that the rate is increasing. But "alarming" is the sort of language that's purely designed to make the story seem sensational, not accurate.

There is, of course, a serious message behind all this. Don't smoke, and don't drink too much alcohol. But I seriously doubt whether sensationalising the statistics in this way is the best way to get the message across.

Update 1.45 pm:

I see that the BBC are now allowing a debate on this story on their Have Your Say site. I find it truly heartening to see that many of the most recommended comments make exactly the same point I made about absolute versus relative risks.