The abstracts I saw presented today tended to have a common theme: they all looked at the risks of therapy with medications we use frequently. You might think that figuring out whether a medication or group of medications cause, say, cancer would just be a matter of following some people and seeing whether they develop the disease. Unfortunately it’s a lot more complicated than that for a few reasons.

For a start, you have to follow A LOT of people for a long time to detect some diseases. When a cancer only occurs in, say, 1 out of 1000 people at baseline over a period of 10 years, you can see that you have to follow a lot of people to see whether it occurs more often in a group of drug users. Seeing 2 out of 1000 people come down with it is a doubling or the risk, so it sounds like that should be straightforward, but how often can you really follow 1000 people for 10 years. And remember, you really have to follow 2000 people because you really also have to follow a group of non-drug users to get real estimates fo the difference in risk: using an historical group of non-drug users isn’t really a good way to do it.

And then, is there really a difference between 1 in 1000 and 2 in 1000? It’s a doubling of risk, but is it just chance – after all it’s only one extra person in a thousand.

It’s also hard because trials of meds rarely use this many people and rarely follow them for a really long period of time. Therefore you have to rely on data that’s really set up for other purposes. For example, you can Medicare and Medicaid claims information with linked prescription data, but that data base isn’t really designed for research and there are a slew of problems with using it.

I feel like I’m starting to teach a course in basic epidemiology, so I’ll stop there, except to say that there is another big problem. It’s well established that many autoimmune diseases like rheumatoid arthritis and lupus increase the risks of some kinds of cancer (especially lymphoma) in and of themselves. And that risk appears to rise depending on how active the disease is in a person. So how do you figure out what’s an effect of the disease, and what’s an effect of the drug used to treat the disease? And when people get some drugs specifically for more active disease, how do you know it’s the drug or the active disease causing the complication?

That’s why there’s a whole group of epidemiologists who look at issues just like this, including your truly, and that’s what this session was mostly about.

The major findings, I think, are related to the TNF blocking drugs that we use to treat some cases of RA. These are the so called ‘biologics’ and include infliximab (Remicaide), etanercept (Embrel), and adalimumab (Humira). Dan Solomon from Boston presented some pretty good evidence that these drugs do not increase the risk of serious infections. Since all of these drugs are thought to suppress the immune system to some extent, and we routinely warn our patients of this risk, this is a good thing to know. It turns out that according to his data, the use of glucocorticoids like prednisone confer a much greater risk of serious infections, and that the higher dose a patient uses, the higher the risk.

I should say that some studies out there suggest just the opposite, i.e. that TNF blockers do indeed confer a greater infection risk, but I’m not convinced that these studies have adequately adjusted the data for the fact that more severe RA in itself also confers a higher risk of infection. They suffer from exactly the problem I described above. In comparison, I think this study did adjust for disease severity sufficiently.

In a similar vein, Fred Wolfe from Witchita presented data that suggest that TNF blockers also do not increase the risk of most cancers (the exception being non-melanoma skin cancers) in patients with RA. I think that prior studies suggesting an association suffer from the same problems discussed above, so that I’m much more inclined to believe these data.

Finally, Dr. Bernatsky suggested that the same thing was true in patients with lupus who are treated with rituximab (Rituxan). I wasn’t quite as impressed with the design of this study as the first two, and indeed she did suggest that the risk of one kind of cancer, non-Hodgkin’s lymphoma, might be higher in patients treated with rituximab. However, this is also the most common cancer associated with lupus, so once again my suspicion is that disease severity was not sufficiently taken into account.

The other interesting study in this session was one given by Dr. De Man which suggested that women with RA do not in general have an inherently higher risk of giving birth to a low birth weight baby, or suffering other pregnancy-related complications. But patients with especially active disease, or those taking prednisone early in their pregnancy did risk a shorter pregnancy and a lower birth weight baby compared to other groups. This is a very good reason to plan pregancies in patients with RA and get disease under control beforehand if at all possible. However, it is reassuring to see that the risk of a low birth weight child is not especially great in general.