First things first. The news that there’s a test for Alzheimer’s which is 90 per cent accurate, even taken completely at face value, does not mean that, if you take the test and it says you’re positive, you’re 90 per cent likely to develop Alzheimer’s disease. Graeme Archer has talked about this stuff in the past, much more knowledgeably than I can, but here’s the basic reason why. (Health warning: this is a discussion of statistics written by someone who hasn’t done maths since GCSE.)
Imagine you test 10,000 people for some disease with a 90-per-cent-accurate diagnostic tool (and for convenience, let’s say that “90 per cent accurate” refers to both the “sensitivity” and the “specificity”, which I’ll explain in a moment). Imagine that it’s a moderately but not hugely rare disease, say 1 in 100 people. Out of your 10,000 people, we can expect 100 people to have the disease.
Your 90-per-cent-accurate diagnostic tool will pick up 90 of those people – only 10 false negatives (that’s the “sensitivity). That’s pretty good; if you’ve got the all clear, it’s 99 per cent likely to be right. But it will also wrongly tell 990 people that they have the disease when they in fact do not – false positives (that’s the “specificity”).
If, then, you take this test, and it says (with 90 per cent accuracy) that you’ve got whatever this disease is, then, far from it being 90 per cent likely that you’ve got the disease, it is in fact only about 9 per cent likely:
(90÷990) x 100=0.0909090…
Alzheimer’s is rather more common than 1 in 100, but even so, without knowing the real figures, you can immediately see that this is not a helpful diagnostic tool; you’d need a test that was significantly more accurate before you can use it to tell people that they’re going to get the disease. As Graeme said earlier today, this finding, if it turns out to be accurate (it’s based on a small trial), will be mainly of use in research: determining whether the blood markers it’s testing are in some way causally involved in Alzheimer’s.
But it has raised the question: would you want to know? If this test was better, or if we were talking about a disease with a more reliable test, you could know years in advance. Huntington’s is the classic; a simple genetic test which will tell you whether you will get the disease, which normally comes on at around 40. But as genetic screening improves, it becomes more and more possible to detect at-risk people. In some cases you can do something about it – certain genes put you at higher risk for certain breast cancers, for instance, and armed with that knowledge people can have preemptive treatment. But in others – Huntington’s, and Alzheimer’s – you can’t, at least not yet.
There’s a logical bit of me which thinks I’d go for it. Say Huntington’s, where if a parent had it, you have a 50 per cent chance of receiving that allele. If you don’t have the test, you still have a 50 per cent chance of contracting the disease if you live long enough, but you have a 100 per cent chance of it hanging over your head. If you have the test, you have a 50 per cent chance of contracting the disease, and a 50 per cent chance of not having to worry about it any more. I don’t like the uncertainty; I’d rather just know.
It’s easy to say, though, when you don’t have to face it right now. Maybe when I’m in my fifties, and a genuinely reliable test for Alzheimer’s becomes available, that’s when the problem will become real.
Tom Chivers 