Why does everything seem to both cause and cure cancer?

Chris put this one to Simon White from the University of Cambridge...

Simon - There are actually very few studies which have shown strong causal links between these factors and cancer - a lot of them are associations. One of the main problems with a disease like cancer, when you’re trying to study its effect when various aspects on the outcomes of cancer, is that cancer is really an older age disease. I generalise there, not all cancers but in broad strokes. So, if we’re looking at people now in their 30s and we say what do you do? We won’t really know if that has an effect until those people are in their 60s, so that would be a causal study. Unfortunately, that takes 30 years and in terms of public health, we kind of need an answer yesterday. So we do a lot more association studies and that’s where some of these competing, contrasting results come from. There’s also a lot of issue with especially diet studies about how we interpret and record diet, so diets have intrinsically changed over the last 50 years. Foods that we have available now in the UK weren’t around 30 years ago. So, can we really say that the impact of various lifestyle factors will cause cancer? - probably not.

Chris - So when people turn round and say the Mediterranean diet, which is dominated by olive oil, a lot of fresh fruit and vegetables, a little bit of red wine every day, that sort of lifestyle and diet combination that will improve your mortality, actually that’s in  the context of people who came from a very different life experience, lived a long while ago really?

Simon - And lived in the Mediterranean.

Chris - And lived in a different geography. We’ve now got people who are different age groups eating a different diet really en mass, and with a different upbringing and, therefore, it may not be relevant?

Simon - And this is one of the issues in statistics is generalising results, so we can’t do the study we want to do. It’s not ethical and it would cost too much money, so we have to work with the information and data we can get and then we have to work out how much that can generalise. This is really one of the great questions of statistics is from the study we have and people we have, what can we say what can we say, what can’t we say? Statisticians, they kind of call us the buzz kills of science because we are usually the ones in the corner going um, is that really true?

Chris - But something like the study done by Richard Dole, who’s died now, but he famously did the smoking doctors study and that was a very well powered study. It had a very big group of people that were followed for 50 years and did categorically show that there was a very big excess of chest diseases and cancers in the individuals who smoked compared with the people who didn’t smoke.

Simon - Yes, and that’s the sort of long term study. It was still an association because he didn’t force certain people to smoke or not smoke.

Chris - He wouldn’t have been allowed to do that though would he?

Simon - No, and that is one of the ethical requirements. To really be true whether something causes cancer or not we have to get a group of people, give it to them, and then another group of people who don’t have it, and then follow them for 30 years.

Chris - Peter?

Peter - Yeah. I think this a really good example with smoking where you have a very binary thing and it’s obviously very bad for you. But it still takes this huge great long 50 year study to work out that smoking is bad for you. Really this is where, in some sense, you can make the case that statistics and these types of studies actually fail public health and public good because they’re not going to determine all of the things in the world that are dangerous for us. Because things where you have a very broad exposure to something, which is not binary and easily definable like whether you smoke or not.  Actually wouldn’t it be fair to say that there could be a vast number of things that we're being exposed to every day which are dangerous to us which we're not being properly… We need a different framework for working out those things rather than these frameworks which can’t pick those signals out from statistical background.

Simon - Just to defend statistics from some comments you made there. Statistics is about understanding what the data can tell you and I don’t think anyone would say that statistics are wrong, they’re a tool.

Chris - As one person said it me the other day that there is reality and you can’t call reality wrong when it doesn’t agree with what you think is going on.

Simon - Statistics is a way of quantifying evidence, and this is where an abuse of statistics occurs when people run these studies and say drinking coffee increases your life expectancy.

Chris - I was really buoyed up when I saw that study. I that that was brilliant - I should live forever the way I’m going.

Simon - But the problem is they've reduced it down to a single aspect - coffee. And we all know that that’s not reality. So the statistics has picked out from all of the various parts of the diet and all of the various lifestyle factors that coffee has this effect, but in combination with a whole host of other things and, yes, these problems are very complex. That doesn’t mean that the statistics is there to tell you actually, we probably don’t know.