Peter D Kramer

In Practice

More Naked Emperors

What if placebos don't work either?

Posted Feb 11, 2009

In the Journal, Robert Temple, of the Food and Drug Administration, examines The Placebo Response and the Power of Unconscious Healing by Richard Kradin, a Harvard-based pathologist and internist who is also a psychoanalyst. Temple reminds that when we refer to placebo controls in scientific trials, we are talking about conditions in which subjects do not get the treatment under study—and to whom a good deal can happen other than a response to the placebo, the "inactive" intervention made to resemble the "active" one. Temple writes, "Patients not receiving treatment may improve for many reasons other than that of having a true placebo response, including the natural history of a disease, other treatment or environmental change, and optimistic assessment of the patient's health by the patient or the physician."

One important and under-discussed explanation for change is regression to (or toward) the mean. In life, extreme conditions—and this observation holds especially for conditions controlled by multiple factors—tend to get less extreme. Temple reminds us that "patients generally enter clinical trials when their symptoms are relatively severe (the entry criteria assure this), so that even for patients with chronic illness, ‘regression to the mean' will, on average, lead to improvement."

One study mentioned in a letter to the Journal in 2001 found that in 30 controlled trials, the apparent placebo effect averaged 9.9 per cent but the expected regressions to the mean ranged from 2 to 37 per cent, so that "the placebo may have no effect, because the regression is large enough to account for the observed improvement after placebo treatment." On its own, regression to the mean can explain the whole of what looks like a placebo response.

This claim—no placebo effect—had been given a strong boost that same year by an instantly-famous study out of Denmark. Asbjorn Hrobjartsson and Peter C. Gotzsche, physician-philosophers at the University of Copenhagen, had looked at 130 trials in which placebos—physical, pharmacologic, or psychological controls—had been compared to "no treatment." The conditions under study ranged from Parkinson's disease to infertility to insomnia, anxiety, depression, phobia, and schizophrenia. Other than perhaps in pain treatment (and in one other specialized research condition) the placebos had no effect at all. Nowhere was their effect in the 35 per cent range that is so often cited.

In response to the Danish article and in support of placebos, a Journal editorialist referred to "that pesky, utterly unscientific feeling that some things just ought to be true." But of course, that same impulse applies to most active treatments under investigation. They are being tested because there is reason to believe that they work—but do they?

In his essay, Temple demonstrates just how hard it is to keep in mind the notion that what we call the placebo response is rarely a response due to placebos. Of the book under review and its author, Temple writes, "Kradin often slips into describing a change in a placebo group as a placebo effect when we do not know that it is truly a placebo effect." Apparently Kadin tries to justify psychotherapy based on its similarity to effective placebos. But Temple concludes, "If the change seen in a placebo group cannot be considered a placebo effect without further evidence, Kradin's attempts to explain that change and characterize it in various settings (which is what most of his book is about), although interesting and in many respects thoughtful, cannot really succeed."

I have not read Kradin's book and so cannot know whether Temple's near-blanket dismissal of it is fair; since the Danish study, theorists have made various interesting efforts to rescue the vanishing placebo response. Still, it is clear that for decades the efficacy of placebos has been overestimated and that in particular this error has marred and continues to mar studies of all sorts in psychiatry. In a future posting, I will explore the implications of this problem. Many ruling arguments in mental health care today are tricked out in references to the placebo response. What happens if we open our eyes and discover that those emperors have no clothes?