Die Another Way

I suppose I will be getting visits from dozens of worried patients who have just read the Herald’s latest attempt to scare the willies out of you.

Popular painkillers ‘lift stroke, heart risk’

Any headline like this, involving common drugs, causes plenty of panic. Many of my patients will simply stop taking these medicines and then visit me when their arthritic pain reaches its peak. Most will be under the impression that just one more Voltaren could kill them stone dead. While this is vaguely true, the risk is no where near as immediate as this article will lead them to believe. It is my opinion that medical reporting of this nature tends to cause a great deal of unnecessary suffering, as frightened patients throw away their tablets and lose confidence in their doctors.

Not that this particular article is especially sensationalist or wildly incorrect. It just lacks a sense of proportion.

For a start, the work in question (PDF here) is not a study per se, but a meta-analysis – a sort of study of studies. Like all of this type of analysis, there are a number of important caveats to consider. The quality of the studies and the quality of each study’s data is very important. A meta-analysis of a dozen poorly constructed studies will produce a poor meta-analysis. Garbage always produces garbage, no matter how elegant the analysis.

The authors themselves point out a significant problem with their study:

“However, the researchers noted that although their analysis covered more than 100,000 patient-years of follow-up, “The number of events for most outcomes was low and our estimates of rate ratios imprecise …””

What they were talking about was the power of each study included in the analysis. Picture a bucket with 10,000 gold coins (I usually find gold coins hold peoples attention while I explain statistics). Let’s say I take 100 coins out and find one is made of brass. I could conclude that there are 100 brass coins somewhere in my bucket – but you can easily see my sample is too small. It may be that there is only one coin, and I was unfortunate enough to find it on my first sample. If there were 20 brass coins, however, you would be more justified in thinking this bucket of gold was riddled with brass. The rate of occurrence of the thing you are measuring, whether brass coins or strokes, is very important. In order to overcome this, you might take a larger sample (say 1000 coins and find 10) or you might do 10 studies of 100 coins and find one in each and add the studies together in a meta-analysis. I’m sure you can envisage that you are more likely to get an odd result in the latter scenario because you might have taken the coins all from the top surface where you could see the brass and pick them out (selection bias) or you might simply be unlucky in your selection methods. Compound this problem by having only 10 brass coins in the bucket and you can see things can get very imprecise.

Okay. That’s enough statistics. I can see your eyes glazing over.

Suffice it to say that this kind of analysis is not very good at separating out various conditions. For instance, is the increase in stokes found in all use of these drugs, or is it only those on long-term or high-dose medicines? Do well people have an increased risk, or is it confined to those at risk already?  We find we can’t answer these questions from this study. We can only make the blanket statement that they may cause an increase risk of strokes and heart attacks. I say may because, in the case of common agent, ibuprofen, for instance, the confidence intervals are wide and the null hypothesis is contained in it. Translation: the range of possible results is large and one of the possible results is that the drug does not increase the risk of strokes and heart attacks.

Basically, we are dealing with a statistical likelihood rather than a gospel truth.

But let us assume that the results are good and accurate and put some clinical perspective on this

Firstly, for most of us, the risk of stroke or heart attack is quite small. In the study, ibuprofen would increase your risk of stroke by three-fold. This sounds like a lot. but let’s say I give you a couple of weeks worth of ibuprofen for your sprained ankle. Your risk of a stroke happening in those two weeks may be as low as 1:30,000, normally – the taking the drug will make it 1:10,000. Small is still small. Obviously longer exposure in a person more at risk makes the risk more significant. That is precisely why we do all these trials – to try and get enough information to make an informed decision as to the use of these drugs. For the most part, the risk appears to be relatively trivial in all but very extended use.

The other perspective is not as encouraging. It is a fact that these drugs can damage the stomach lining, causing ulcers and can also damage your kidneys. In fact, it is far more likely that you will experience one of these side effects, than you will suffer a stroke or heart attack – particularly if you use these drugs for extended periods of time. Most people who are on long-term anti-inflammatory drugs will be aware of this and their doctor will have discussed it with them. The point here is that this is a much more significant risk, because it is much more common. People who are on these drugs long-term are therefore already taking a much greater risk than one of increased heart attacks or strokes. Consequently, few people will need to stop these drugs for the small increase in risk that this study reveals.

Your next Voltaren is probably not going to kill you, but untreated arthritis may well make you wish you were dead…