Proof, puddings and surrogate outcomes

TW Collins' Bread Pudding

The proof of the pudding is in the eating, we’re told. Well, yes. But what if someone told you that, actually, how the pudding looked before it went into the oven was enough to go on, or perhaps how it smelled while cooking? They might be a good guide to whether the pudding would turn out well, sure, but you still want to know what it tastes like.

The equivalent in medical trials is the ‘surrogate outcome’. For example, if you want to know whether a drug prevents people dying from a heart attack, you need a pretty long trial. It may take 10 years before you have enough data to know whether the drug works, because people don’t die off straight away. A 10-year trial means a lot of invested money for the drug company, possibly with a negative result at the end of it. No wonder that drug companies want quicker results.

Which is why so many trials look at surrogate outcomes. So they’ll say, well, we know that taking statins to lower cholesterol prevents heart attacks. That means if we can show that another drug lowers cholesterol (which we can show in a matter of months) , that should prevent heart attacks too.

Or you could look at whether a drug can increase bone mineral density – the proportion of calcium and other minerals in your bones – because a denser bones are less likely to break, and so this drug should prevent fractures.

Another example is in diabetes. High blood sugar damages blood vessels. Damanged blood vessels can cause heart attacks, so a drug that can control blood sugar should prevent heart attacks.

That, in a nutshell, is how three recent drugs got licensed to prevent heart attacks, fractures and complications from diabetes. The trouble is, the long-term data suggests they may not actually work as expected.

The first, ezetimibe, works really well to prevent the body from absorbing cholesterol from the gut into the bloodstream. It lowers levels of ‘bad’ (LDL) cholesterol. But it doesn’t seem to prevent the build-up of fatty plaques in arteries, and there’s no data yet about whether it can prevent heart attacks.

The second type of drug, called bisphosphenates, work both to improve bone mineral density, and to reduce the chance of fractures, for people with osteoporosis (weak bones). The controversy is around whether women who don’t have osteoporosis, but are at risk of getting it in future, might benefit. Many women get weaker bones in older age. But bisphosphonates can have unpleasant side effects, and seem to increase the risk of cancer of the oesophagus. So we need to be pretty sure that they really do prevent fractures, before they’re more widely used.

The third drug, rosiglitazone for diabetes, certainly works to lower blood glucose. But instead of reducing the risk of heart attacks, long-term data now emerging shows it actually increases the risk of heart problems for some people. There have been recent calls for it to be withdrawn from use in the UK, with drug advisors saying it has no place on the UK market. A far cry from its introduction in 2000, when I remember excited PR companies extolling its virtues as the next big thing in diabetes care.

So we have a dilemma. Do we wait 10 years for ‘proper’ data about whether a drug does what you want it to? In which case, some protest, the drugs companies will give up and patients will be deprived of effective treatments. Or do we go along with the smell of the pudding in the oven, only turning off the gas when the smell starts to get distinctly fishy?

Image: From TW Collins’s photostream, with CCL, on Flickr

UPDATE 28 September 2010: The European Medicines Agency last week withdrew marketing authorisation from rosiglitazone, saying that the benefits of the drug no longer outweigh it’s benefits.

%d bloggers like this: