Weak Drugs or Powerful Placebos? One Side of the Issue

Steve Silberman at Wired magazine has written a truly phenomenal paper about drug manufacturers and the placebo effect in research:

http://www.highbeam.com/doc/1P2-6724828.html

It's right a good deal of the time, and where it's wrong, it's wrong in a fascinating way. I expect to get several posts out of Silberman's article, to keep each within manageable length.

The subject of this post is a "pro-industry" assessment of Silberman's main claim--that somehow, in recent years, the placebo effect in drug research has gotten more powerful, making it harder for the companies to prove that even good drugs are truly superior to placebo (the minimum bar for FDA approval in most cases). The next post will be the skeptical assessment of the same claim.

The news flash that Silberman passes on--citing the work of psychiatrist and Merck VP William Potter--was actually something we had a pretty good glimmer of a quarter-century ago, from a classic placebo paper by anthropologist Dan Moerman. Moerman did a meta-analysis (actually one of the very first I think) of 31 studies of cimetidine vs. placebo. (Anyone remember Tagamet? Before the "purple pill" era of Prilosec and its lookalikes, cimetidine was the real breakthrough drug, turning peptic ulcers from a surgical into a medically treatable condition.) Moerman showed that on average, the rate of good therapeutic responses among those given placebos for their ulcer disease, in the context of these double-blind studies (with ulcer size being measured by endoscope visualization, so we are talking here about "objective" responses, not merely "subjective" relief of symptoms) was about one-third. That was sort of reassuring as it matched the estimate given by Henry Beecher in his previously classic, but now widely debunked paper, "The Powerful Placebo" published in 1955.

Moerman proceeded to show that the one-third figure was more obscuring than enlightening, because the main phenomenon about placebo response rates were their wide variability. The placebo response in one study hit a low of 10 percent, and in another study topped out at 80 percent. If you looked at all 31 studies, about half showed cimetidine being superior to placebo, and half not. The cimetidine response rate was actually very uniform--about 75 percent of subjects improved in the drug group in nearly every study. So whether cimetidine was superior to placebo or not said nothing about how cimetidine performed; it was all about how well placebo performed.

(Sidebar to the main story line: The astute reader will already be jumping up and down, that I have confused two quite different phenomena--the true placebo response rate, which is the healing that occurs because you take a placebo; and the natural history of peptic ulcer disease, which says that if you wait long enough, most ulcers will heal on their own, even if you took nothing at all. The placebo group in a typical double-blind controlled trial is subject to both effects, and the trial methodology does not distinguish which is which. But this was one of the significant points about what Moreman found. Can you really argue that peptic ulcers heal by themselves only 10 percent of the time in one place, but 80 percent of the time in another? The wide variability among study sites and contexts strongly suggested that it was most likely the "true placebo effect" that varied from study to study, while the "natural history effect" would be expected to be present, but not to be so widely variable.)

You see right away the implications for the drug company. Imagine that through bad karma, all 15-16 or however many studies that happened to show cimetidine not being any better than placebo, had been the first set of studies performed. The companies would never have taken the drug to the FDA for approval. And both in those studies and also based on extensive clinical and research experience subsequently, cimetidine is as near as you are going to get to a really good drug. Indeed, we now think that among all the patients being prescribed expensive proton-pump-inhibitor drugs like Nexium, the majority would probably have gotten just as much better, at far lower cost, on drugs of the cimetidine class.

Silberman's article takes this story up to date by discussing recent research, especially in psychiatric conditions like depression, that seem to indicate that there has been a steady upward placebo creep in recent years, so that it is getting harder and harder to show that even supposedly good new antidepressants are better than placebo. Why this might be happening is the mystery the article delves into. Potter years ago suggested that the problem was so big that it would behoove the drug companies actually to join forces and review their pooled data sets to try to crack the puzzle and figure out how better to test their drugs so that they stood a fighting chance against placebo. For years the companies said, no way will we share our data with our competition. Finally they have seen the wisdom in Potter's idea and NIH has gotten involved as well. Here's Silberman's summary:

"Under the auspices of the NIH, Potter and his colleagues are acquiring decades of trial data—including blood and DNA samples—to determine which variables are responsible for the apparent rise in the placebo effect. Merck, Lilly, Pfizer, AstraZeneca, GlaxoSmithKline, Sanofi-Aventis, Johnson & Johnson, and other major firms are funding the study, and the process of scrubbing volunteers' names and other personal information from the database is about to begin.
In typically secretive industry fashion, the existence of the project itself is being kept under wraps. NIH staffers are willing to talk about it only anonymously, concerned about offending the companies paying for it."

(Now, anything this secretive has its own set of problems which we will have to address in a later post.)

Silberman tosses around a few preliminary hypotheses about why placebos supposedly work so much better these days than before, of which the most interesting and ironic one is that the companies have poisoned their own well with their massive DTC advertising campaigns. Data suggest that placebo response is commonly much higher in developed countries than when the same studies are done in poorer nations. And of course the concentration of prescription drug advertising, brainwashing the viewer into thinking that a pill conquers every ill, is much greater in the former nations. In a picturesque phrase, Silberman summarizes, "one way that placebo aids recovery is by hacking the mind's ability to predict the future." Heavy doses of drug advertising means that the brain predicts the future in a way that maximizes reaction to placebos.

That's one way to spin all this. Next I'll look at it from a more skeptical direction.

Moerman DE. General medical effectiveness and human biology: placebo effects in the treatment of ulcer disease. Med Anthropol Q 14(4):3-16, 1983.

Beecher HK. The powerful placebo. JAMA 159:1602-6, 1955.