Statistics has generally been a weakness in my mathematical armoury. This is in most part because it is quirky, and frequently counter-intuitive.
It is for these very reasons that when a health message is predicated on a statistical feature, you really want to be sure that the messenger is being honest with you.
The recent Lancet report of cholesterol used a meta-analysis of prior research results to conclude that all over-50s should be given statins even if they were healthy as a preventative treatment. The matter of side-effects, which can be pretty nasty for statins, was downplayed.
The conclusion was that you could reduce the long term chance of a coronary heart disease (chd) by a certain percentage by taking them. The exact details are not so crucial here as the principle in such statistical matters is that they sound appealing more because of the manner of expression than the reality that this expression hides.
If the chance of something happening is currently 5% for most people, then it will happen to 5 in 100 people. If action is taken to reduce that chance by 20%, then the 5% is reduced to 4% or 4 people in 100.
If all 100 of those people had to take a tablet every day for the rest of their life to try to appreciate the advantage of the 20% improvement, then only 1 of their number will gain, statistically speaking. So 99 people will be taking the tablet for no real gain – 99% of people doing something for no reason. This is a much more telling statistic.
But the large percentage now recommended to take statins will not just be wasting their time, but now introducing ‘side-effects’ into their bodies, adding to risks of other ailments.
This is why stats can be very dangerous – even a reasonable reduction in a small probability does not make much difference. By way of example, if you buy a second lottery ticket, you increase your chances of winning by 100% but the chance is still in the millions to one against.