Got Placebo?

It was just over 3 years ago, and reporters breathlessly heralded the news: "St. John's Wort Is Ineffective For Depression". Unfortunately, they got the story wrong, and missed the real story sitting right under their noses.

Here's what happened. The National Institutes of Health had just sponsored a large outcome study in which hundreds of depressed patients were randomly assigned to one of three treatment conditions: Zoloft (the best-selling antidepressant), St. John's Wort, or Placebo (an inert sugar pill). None of the treatments worked particularly well. After 8 weeks, the following proportions of patients were found to be recovered within each treatment group:

Placebo - 32%
Zoloft - 25%
St. John's Wort - 24%

Notice anything interesting? Yes, oddly enough, the sugar pill yielded the best results of all.(Statistically speaking, though, the 3 treatments were judged to be in a virtual tie.) Now, in order for medical researchers to conclude that any given drug is effective, it has to outperform a placebo control condition. Clearly, this didn't happen for St. John's Wort (hence the headlines) . . . but it didn't happen for Zoloft either (a fact that was completely ignored by the press)!

However, at this point, perhaps you're thinking, "Surely this study is some sort of anomaly. A fluke outcome. We already know that Zoloft and similar drugs are much more effective than a sugar pill." But do we?

Clinical researcher Irving Kirsch and his colleagues recently petitioned the Food and Drug Administration under the Freedom of Information Act for data on the 47 drug trials submitted by the pharmaceutical industry in their quest to get FDA approval for 6 of the most popular antidepressant medications (Prozac, Paxil, Zoloft, Celexa, Effexor, and Serzone). Most of these drug studies had never even been published, as drug companies tend only to publish the studies that show their drugs in a favorable light . . . but the FDA keeps permanent records of every such trial.

What did the FDA records show? Remarkably, Kirsch and colleagues found that in the majority of these drug trials, the antidepressant did not beat the placebo. In fact, when the researchers averaged across all 47 studies, they found that the placebo led to symptom reduction that was fully 85% as large as that of the active medications. This amounted to a 2-point average difference on a 52-point symptom rating scale; this magnitude of difference is not considered to be clinically significant.

Does this mean that Zoloft and similar drugs don't work? No. It's clear that they lead to complete remission for about 1/3 of the patients who take them (and to some improvement for many others) . . . It's just that much of this benefit for many patients is based on the placebo effect. The data on this point are crystal clear.

And how does the placebo effect work? This will be the subject of an upcoming post . . .