Sunday, 15 February 2015

Is cancer really just due to bad luck?

Consider this press release about a potentially fatal disease. It was put out last month by one of the world's top universities - Johns Hopkins in the US - and it basically says that two thirds of cancer risk is down to random mutations in your genes. There's nothing you can do about it. That's just life - analogous to a car journey or a train trip. There might be an accident. That's unfortunate, but you might as well worry about getting hit by lightning.

Except that the whole basis of the story was faulty. That wasn't the fault of the researchers, but of the university press office that put out the fatal phrase 'two thirds of cancer just bad luck'. Er, no. That's not supported by other science, actually, and Cancer Research UK estimates that 40% of cancers could be prevented by changing our lifestyles.

Why the difference in numbers? Well, they're down to the fact that the Johns Hopkins scientists were measuring the amount of variations in cell division and their relationship with numbers of cancers - the more cells split up, the more risk of cancer (above). So far, so good. But that didn't have a baseline. It didn't start from the actual level of cancer risks, to you or anyone else (or in the population as a whole). It dealt only with the variance - the amount of change explained by the cells' behaviour. So the amount these ecological experiments tell you about the actual level of the cancer risk? Not that much. Not only that, but the confidence intervals of the research (the level at which the researchers can be 95% certain that their result is 'correct'), was actually 39% to 81%. That's a massive range which allows us to speculate a lot, even about the narrow question of cell division and the specific links between changes in that process and the potential prevalence of cancers. And the analysis gives a lot more weight to relatively rare types of cancer (including relatively unusual inherited forms) than it does to the ones you should actually worry about dying from.

So the answer to the question with which we opened is just: no, we can't say that most cancers are just due to bad luck. This research does not say that two thirds of actual cases, and certainly not each individual's risk of cancer, add up, sixty-six times out of a hundred, to pure chance. Most of it probably does happen via processes that look like chance to us, but in interaction with the environment. It's not, as any scientist will tell you, an 'either/ or' question. Go back to that driving analogy. Yes, there's always a risk of an accident. But if you want to increase that risk, you can drive erratically, drunk, badly or without due care and attention. That'll change the parameters. If you want our advice, you should eat lots of fruit and vegetables (and wholemeal products), get out on brisk walks, limit your alcohol intake, make sure you don't smoke, and watch your weight. In fact, exactly what doctors and statisticians have been telling you for years.

The whole imbroglio is deeply worrying. But we can't just blame journalists - or even university press officers, who are usually much more up to the mark. They only write stuff like this because they think that's what the audience wants - and what they must pander to in order to get a hearing. There's a lot of poorly-orientated numerical thinking out there, from the clangers put out by anti-vaccination activists to most people's faulty knowledge of economics.

As we've said again and again on this blog, statistics are constructed - they happen in the relationships between the makers and the users (or interpreters). Like class, which happens as the machinery of voice and language and prejudice moves - as a verb, rather than a noun, as it were - so statistics are produced, understood (or misunderstood), read, re-read, handed on, garbled or sharpened. They happen in the space and the relationship between the citizen and the scientists or social scientists involved. So we're part of it all. We're implicated in the successes and the mess-ups. We each have a responsibility to tool ourselves up in the maze of quantitative choice and the mathematical fog in which we find ourselves.

Don't be put off by bad press releases, media units' indiscriminate use of data, dodgy numbers and poor reporting. It's rampant - as this unfortunate example demonstrates. Be confident. Equip yourself with the statistical, numerate, quantitative skills of scale and scope that will allow you to fight back - an ability to interpret data that is becoming more and more important in our statistical age.

Healthy scepticism is a very important part of any functioning democracy. But we should all be able to extend it to those questioning the intellectual or scientific status quo, just as we direct it at our governors and leaders.

No comments:

Post a Comment