Do physicians really believe in placebos?

This article is cross-posted at Science-Based Medicine. Check it out. –PalMD

ResearchBlogging.orgIn a previous post, I argued that placebo is an artifact of certain clinical interactions, rather than a treatment that we can exploit. Apparently, there are a whole lot of doctors out there who don’t agree with me. Or are there?

A recent study published in the British Medical Journal is getting
a lot of enk (e-ink) in the blogosphere. As a practicing internist, I have some pretty strong opinions (based in fact, of course) about both this study and placebos in general.

The Study

The current BMJ study defines placebo as “positive clinical outcomes caused by a treatment that is not attributable to its known physical properties or mechanism of action.” I’ve got a lot of problems with this definition, but we’ll get to that later. It also allowed physiologically active medications to “count” as placebos. Oops.

The study surveyed internists and rheumatologists practicing in the U.S. They tried to control negative responses to the term “placebo” thusly (from the Methods section):

Because the term “placebo” and behaviours surrounding its use can be contentious, we devised a series of non-judgmental questions beginning with broad questions that avoided the term “placebo” and then gradually gained more specificity, culminating in items whose responses used a clear definition of a “placebo treatment.” By constructing a series of items in this manner we allowed respondents to describe their attitudes and experiences as accurately as possible.

The first set of three items began with a hypothetical scenario in which a dextrose tablet was shown in clinical trials to be superior to a no treatment control group (thus establishing its efficacy as a placebo treatment). To avoid biasing responses these three questions did not use the term “placebo,” “placebo treatment,” or “placebo effect.”

I know a lot of doctors. They tend to be fairly bright. I don’t think any of them would be deceived by this scenario. When you read “dextrose pill”, you think “placebo”. It might as well be the same word.

Respondents were then asked to indicate which of several treatments they had used within the past year primarily as a placebo treatment, defined as a treatment whose benefits derive from positive patient expectations and not from the physiological mechanism of the treatment itself; and how they typically described placebo treatments to patients. By asking these five questions both without the term “placebo” and then using the term, we aimed to assess physicians’ practices as accurately as possible.

All docs would recognize this as a placebo scenario. However, the definition above is rather problematic. Is a placebo “a treatment whose benefits derive from positive patient expectations and not from the physiologic mechanism of the treatment itself”?

The validity of this study hinges on the answer to this question.

Placebo—I do not think it means what you think it means

The concept of placebo, and the way it is used in this study are both problematic. First, we have an elephant in the exam room. When we observe a so-called placebo effect, we are very susceptible to the post hoc ergo propter hoc fallacy. Just because the patient’s condition changes doesn’t mean we have done anything to cause that change. In fact, due to the remarkably inexact human pattern-recognition software, we are likely to attribute a change in a patient’s condition to something, and if we don’t know what that something is, we may label it “placebo”. So the very concept of placebo may be an artifact of our way of thinking, a label to place on a co-incidence, rather than a “thing”. We may have wrongly reified a rather fuzzy concept.

So, I’m not quite sure what a placebo is. In the current study, a placebo is defined as “positive clinical outcomes caused by a treatment that is not attributable to its known physical properties or mechanism of action.” This implies that the physician either knows the treatment shouldn’t work, or doesn’t understand how it works. This isn’t just semantics; we have many treatments available whose exact mechanism of action isn’t known, but whose effectiveness has been proved. If you interpret the definition less strictly, it oxymoronically defines a placebo as something that works despite it’s lack of efficacy. If I prescribe something expecting a predictable effect, and it produces that effect, by definition it isn’t a placebo. If I prescribe something I expect to work, and it doesn’t, then it isn’t a placebo. If I prescribe something expecting failure, but it works, I’m a lucky idiot. This would seem to imply that there is no such thing as a placebo (and I might agree).

Earlier work lays down some less problematic definitions by dividing out the “components” of the placebo effect. One scheme divides placebo effects into effects due to assessment and observation, due to therapeutic ritual (placebo treatment), and due to supportive patient-practitioner relationship. (It leaves out coincidence.) Each of these effects is easy to remove from the shadow of placebo. Effects due to assessment and observation occur with all patients. Therapeutic rituals exist independent of traditional placebos—for example, simply doing a thorough exam is a useful ritual, and is often therapeutic. And a supportive doctor-patient relationship is (should be) standard. When you try to divide placebo into its components, you’re left with standard components of medical care.

Modern medicine has a bit of a Clarkian quality to it. Most patients don’t know how medicine works, and that’s OK. It’s not necessary that every patient be a trained physician. Given the power of modern, science-based medicine, it can seem like magic to the uninitiated. To someone who doesn’t apply too much thought to the problem, a placebo can seem to be another tool on the shelf along with acetaminophen, appendectomies, and atenolol. Each of these tools can have a positive clinical effect. And positive effects can occur with no treatment at all. Is it right to label that which we do not understand “placebo”?

The best definition of placebo that I’ve seen, snatched right from Wikipedia, is ” a substance or procedure a patient accepts as medicine or therapy, but which has no specific therapeutic activity.” This seems more accurate. It takes out the necessity for a placebo to actually work. Some might say that this eviscerates the concept entirely, and that’s really the point—a placebo treatment is not a treatment at all, only the perception of one.

In sum…

A placebo is a tool used in clinical studies to separate out effects of the treatment in question from doing nothing. If there are improvements in subjects in the placebo arm, it is usually called a “placebo effect”, but may just be either random changes in the patient’s condition, or the effect of standard medical care such as being observed and cared for. A placebo cannot be used as a treatment, and cannot be observed outside of this setting. After all, how would you test a placebo? Against a different placebo? And trials that do exist measure only subjective outcomes, such as pain, rather than something objective, such as tumor size. Would you be willing to be a subject in a trial of placebo vs. nothing at all for your colon cancer? The ethical problem of deceiving a patient in order to achieve a placebo effect is dealt with in detail in Dr. Gorski’s post from earlier today. I would add to his argument that since placebo effects don’t really exist outside of clinical studies, there is never a good reason to use one.

Finally, placebo is a refuge for cult medicine enthusiasts. When homeopaths, naturopaths, chiropractors, and other modern shamans are confronted with the utter implausibility of their offerings, they may retort, “well, it at least has a powerful placebo effect, and the patient feels better.”

A careful examination of the nature of placebo shows this argument to be bankrupt. Placebo is not a treatment. It is unethical. And, as it is commonly understood, placebo treatments may in fact be non-existent, an ephemeral change in a patient’s condition. To claim credit for it is disingenuous at best.


J. C Tilburt, E. J Emanuel, T. J Kaptchuk, F. A Curlin, F. G Miller (2008). Prescribing “placebo treatments”: results of national survey of US internists and rheumatologists BMJ, 337 (oct23 2) DOI: 10.1136/bmj.a1938

T. J Kaptchuk, J. M Kelley, L. A Conboy, R. B Davis, C. E Kerr, E. E Jacobson, I. Kirsch, R. N Schyner, B. H. Nam, L. T Nguyen, M. Park, A. L Rivers, C. McManus, E. Kokkotou, D. A Drossman, P. Goldman, A. J Lembo (2008). Components of placebo effect: randomised controlled trial in patients with irritable bowel syndrome BMJ, 336 (7651), 999-1003 DOI: 10.1136/bmj.39524.439618.25

A Hróbjartsson (2002). What are the main methodological problems in the estimation of placebo effects? Journal of Clinical Epidemiology, 55 (5), 430-435 DOI: 10.1016/S0895-4356(01)00496-6