Sunday, August 23, 2009

Of Carts and Horses

Last week, I wrote about a paper finding that the mosquito repellent chemical, DEET, inhibits an important enzyme, cholinesterase. If DEET were toxic to humans, this finding might explain why.
But it isn't - tens of millions of people use DEET safely every year, and there's no reason to think that it is dangerous unless it's used completely inappropriately. That didn't stop this laboratory finding being widely reported as a cause for concern about the safety of DEET.

This is putting the cart before the horse. If you know that something happens, then it's appropriate to search for an explanation for it. If you have a phenemonon, then there must be a mechanism by which it occurs.

But this doesn't work in reverse: just because you have a plausible mechanism by which something could happen, doesn't mean that it does in fact happen. This is because there are always other mechanisms at work which you may not know about. And the effect of your mechanism may be trivial by comparison.

Caffeine can damage DNA under some conditions. Other things which damage DNA, like radiation, can cause cancer. But the clinical evidence is that, if anything, drinking coffee may protect against some kinds of cancer (previous post). There's a plausible mechanism by which coffee could cause cancer, but it doesn't.

Medicine has learned the hard way that while understanding mechanisms is important, it's no substitute for clinical trials. The whole philosophy of evidence-based medicine is that treatments should only be used when there is clinical evidence that they do in fact work.

Unfortunately, in other fields, the horse routinely finds itself behind the cart. An awful lot - perhaps most - of political debate consists of saying that if you do X, Y will happen, through some mechanism. If you legalize heroin, people will take more of it, because it'll be more available and cheaper. If you privatize public services, they'll improve, because competition will ensure that only the most efficient services survive. If you topple this dictator, the country will become a peaceful democracy, because people like peace and democracy. And so on.

These kinds of arguments sound good. And they invite opponents to respond in kind: actually, legalizing heroin is a good idea, because it will make taking it much safer by eliminating impurities and infections... And so the debate becomes a case of fantasizing about things that might happen, with the winner being the person whose fantasy sounds best.

If you want to know what will happen when you implement some policy, the only way of knowing is to look at other countries or other places which have already done it. If no-one else has ever done it, you are making a leap into the unknown. This is not necessarily a bad thing - there's a first time for everything. But it means that "We don't know" should be heard much more often in politics.

No comments:

Post a Comment