The placebo effect is a real problem in some clinical trials. It varies, but in things like antidepressants it's a major factor (while with, say, pancreatic cancer it doesn't change the results too much.) In a given sample of depressed patients, there are a fair number of people (20 or thirty percent) who will respond if you give them 50 milligrams of confectioner's sugar which they truly believe to be an efficacious drug.

Of course, the majority will respond as if you'd given them, well, confectioner's sugar, but that group of placebo responders will blow your statistical workup to pieces. This is one of the reasons that you see multiple trials for antidepressants, because the trials themselves often just produce noisy data. Of course, one way to interpret this is that the antidepressants themselves are fairly worthless. That's a tempting conclusion, and for some people, they clearly don't do much good. But you can find others that truly appear to have been helped. Depressed patients, even ones who may look and act similarly, are clearly a heterogeneous population.

What if those strong placebo-responders could be weeded out of the patient population before you even started the clinical trial? This question is a good test of a person's attitude toward the drug industry. Many folks will hear that idea and cry "Fraud! Stacking the deck!" But think about it. If you could find the people who will improve when given a sugar pill, then you could pull them aside and just go ahead and give 'em the sugar pill. Hey, it's effective therapy, and that's what counts, right? And they'll miss out on the side effects of the antidepressant drugs themselves, and every drug has side effects at some level - every single one.

Meanwhile, once those folks have been sorted out, you're left with a cohort of patients who need all the help they can get, and now you're in a statistical position to see if you can really provide any. As far as I can see, everyone comes out ahead.

It turns out that there may be ways to see who's a strong placebo effect candidate and who isn't. There have been several studies in the last few years that show some real correlations in brain activity during placebo situations, and this has lead to the idea of a test for it.

If this goes on, though, there could be some interesting developments. What if everyone becomes aware of the test to see if you're going to get a placebo? Will the responders still respond if they thing there's a reasonable chance that they didn't get a "real" drug? I think that what we'll need to do is present the test as a standard procedure, to help figure out which therapy would work the best - not a method to see if you get a drug or not, but a method to see which drug you should get. That should keep things working.