Map of life expectancy at birth from Global Education Project.

Tuesday, July 25, 2006

Just say no?

We can't, and shouldn't, just say no to drugs. It makes a lot of sense to take some drugs, under some circumstances. Surgery without antibiotics, for example, would be a very bad idea. So would treating pneumonia with chicken soup. It is pretty clear that for people who have heart disease and have had coronary events, taking drugs such as statins and ACE inhibitors prolongs life, and it is at least a reasonable decision for people who have established risk factors for heart disease to take such drugs. And so forth.

However, it is often -- no, I'll say usually -- the case that sorting out whether it is in a particular person's interest to take a particular drug at a particular time is very difficult. It may well be a matter of judgment with no clear answer. It's often a bad idea, even a very bad idea, but it's easy for prescriber and patient not to know it. Here are some of the reasons.

First of all, drug trials are designed to show that drugs do good; they generally aren't designed to detect or establish harmful effects. There are at least four problems here, two of which are fundamental philosophical difficulties (FPDs).

  1. Since we can't generally predict in advance what harms might be caused by a drug, we don't even collect data on all possible adverse effects during a clinical trial. We couldn't if we wanted to, since the possibilities are unlimited. We will keep track of deaths and events of major clinical significance, but we can't measure every possible biological parameter that might be adverse, but doesn't produce immediately noticeable symptoms.
  2. Clinical trials used to establish safety and efficacy usually don't go on for very long -- seldom more than six months. It often takes longer than that for adverse effects to emerge.
  3. Here's an FPD: tests of statistical significance depend on specifying end points in advance. We want to prove that Rebleckafecklameckla* lowers blood pressure, and we note a difference between the treatment group and the control group which is too large to have arisen by chance. There are bound to be innumerable other differences between the two groups, just by coincidence, but we can't establish p values for them because if you make ten or more comparisons, you're likely to get at least one spurious association.
  4. Here's another. The study is sufficiently "powered" - i.e. has enough participants -- to detect the clinically important effect we're looking for. But there might be adverse effects which are relatively uncommon, which the study is too small to detect, but which are serious enough to outweight the benefits, even if they are rare. I.e., Vioxx relieves pain very reliably, but so do lots of drugs. It also occasionally kills you, which many of the alternatives are less likely to do, but the study was too small to prove that. (Actually it wasn't, but the drug company didn't want to interpret it that way. Which is another problem, I suppose.)


So, evidence of adverse effects typically emerges only after a drug is approved and used in thousands or millions of people. But then we aren't doing controlled studies of its effects. If we start to get reports that people taking Rebleckafecklameckla are developing excessive nose hair, is that really an effect of the drug or are people just attributing it to the drug because they happen to be taking it?

But it gets even more complicated. Drugs may be risky for some people but not others, depending on factors such as what other drugs you're already taking and any number of medical conditions. With 7,000 drugs the number of possible interactions is astronomical. Doctors can't possibly keep track of all the counterindications, cautions and interactions, and for patients to truly make informed choice about taking drugs is, quite frankly, a totally unrealistic goal. And, of course, what we know about adverse effects, counterindications and interactions continually changes, long after drugs and their "labels" (really novelette length essays written in highly technical language) are approved, while it can take years to update the labels.

And, of course, drugs can be very expensive.

So what's a person to do? I say, the best you can do is to be conservative about taking medications -- if not doing it seems like a reasonable choice, it's probably the right one. Ask questions. Read that damn package insert. Check out the info from Consumer Reports on the "best" drugs, or, for a more negative view, check out what Public Citizen has to say about the worst drugs. Of course, a lot of drugs get both ratings. Hmm.

Whatever you do, don't depend on the FDA to protect you. They aren't working for you.

*Obviously I made up the name but they are starting to run out of decent names for drugs, as you may have noticed from their increasing bizarreness.

No comments: