What's the difference between relative and absolute risk?

Chris asks statistician Simon White...

Chris - You see this in the newspaper quite a bit. You see people reporting on a science or a medical study and they say the relative risk in this group was X compared to people who didn’t smoke, or drink, or drive very fast the wrong way down the motorway or something. What does that terminology actually mean though?

Simon -  This is a really good question and it’s one of these areas of science that over the years has become a very technical definition of risk. In its essence, absolute risk is the kind of risk we want to normally  talk about - how many out of 1,000 people will die from say lung cancer? That’s what we want to know as a wider public when we’re thinking about where we want to prioritise certain conditions; whether we should be worried about certain things. However, most studies are done by comparing different types of patient groups across different demographics: age, sex. At which point you then make relative comparisons between those groups, so it’s how much worse is this for men than say for women? So you get a relative risk from the study but that’s not always the number that you want or is the best number to make it helpful to understand, that’s usually the absolute risk.

Chris - And that’s the distinction?

Simon - Yes. It’s about what those numbers are in relation to. So an absolute risk is an actual countable number of people. A relative risk makes no sense on it’s own - it’s only relative to something else.

Chris - But if you’ve done your study carefully and you’ve got, for instance, a group of people who you didn’t do anything to, you just watched them, and then you have a group who are analogous. They’re identical to that group as far as you know to the greatest extent possible, they eat the same things, they exercise the same amounts, the same body weight all that, and they have the same sort of genetic background. You do something to this other group - you’re comparing, you hope, apples with apples and so that relative risk is actually very informative there isn’t it if the second group that you do something to have a much lower chance of developing cancer or heart disease or something, that relative risk in the second group is informative?

Simon - Absolutely. A lot of the key results we know that inform public health and medicines that we use every day are based on relative risks from these kind of studies, so they’re not useless, they’re just a different way to communicate that information. And, in fact, the debate about how best to communicate risks is a longstanding challenge in statistics.

Chris - I was going to say, it’s not something that the general public often get told how to interpret. They just see all these facts and figures reported in various media and no one ever really explains to you how you should interpret or understand a lot of those points. And this, I think, is why people get very confused when they see one week a news story says “chocolate will protect you from dying of high blood pressure”, and then the next week it says “chocolate makes you fat and gives you high blood pressure” - then what should I believe?

Simon - Again, this question of causation and association. A lot of studies can’t actually definitively prove that one thing leads to another. A lot of them are certain kinds of observational study so we get these relative risks but they’re not always set in stone.