The Dece(i)bo Effect

The Dece(i)bo Effect — think portmanteau of Deceive and Placebo — refers to the facile application of constructs, without unpackaging the concept and the assumptions on which it relies, in a fashion that, rather than benefiting thinking, leads reasoning astray.  

Words and phrases enter common parlance, that capture a concept: Occam's razor, placebo, Hawthorne effect. Such phrases and code-words in principle facilitate discourse — and can indeed do so. Deploying the word or catchphrase adds efficiency to the interchange, by obviating the need for pesky review of the principles and assumptions encapsulated in the word.

Unfortunately, bypassing the need to articulate the conditions and assumptions on which validity of the construct rests, may lead to bypassing consideration of whether these conditions and assumptions legitimately apply. Use of the term can then, far from fostering sound discourse, serve to undermine it.

Take, for example, the "placebo," and "placebo effects." Unpackaging the terms, a "placebo" is in principle something that is physiologically "inert" — but believed by the recipient to be active, or possibly so. The term "placebo effect" refers to improvement of a condition when persons have been placed on a placebo, due to effects of expectation/suggestion.

With these terms well ensconced in the vernacular, Dece(i)bo Effects associated with them are much in evidence.  Key presumptions regarding placebos and placebo effects are more typically wrong than not.

1. When hearing the word "placebo," scientists often presume "inert" - without stopping to ask: what is that allegedly physiologically inert substance? Indeed, even in principle, what could it be??

There isn't anything known to be physiologically inert. There are no regulations about what constitute placebos; and their composition — commonly determined by the manufacturer of the drug under study — is typically undisclosed.  Among the uncommon cases where placebo composition has been noted, there are documented instances in which the placebo composition apparently produced spurious effects. Two studies used corn oil and olive oil placebos for cholesterol-lowering drugs: one noted that the "unexpectedly" low rate of heart attacks in the control group may have contributed to failure to see a benefit from the cholesterol drug. Another study noted "unexpected" benefit of a drug to gastrointestinal symptoms in cancer patients. But cancer patients bear increased likelihood of lactose intolerance — and the placebo was lactose, a "sugar pill." When the term "placebo" substitutes for actual ingredients, any thinking about how the composition of the control agent may have influenced the study is circumvented.

2. Because there are many settings in which persons with a problem, given placebo, report sizeable improvement on average when they are re-queried (see 3), many scientists have accepted that "placebo effects" — of suggestion — are both large in magnitude and widespread in the scope of what they benefit.

The Danish researcher Asbjørn Hróbjartsson conducted a systematic review of studies that compared a placebo to no treatment. He found that the placebo generally does: nothing. In most instances, there is no placebo effect. Mild "placebo effects" are seen, in the short term, for pain and anxiety. Placebo effects for pain are reported to be blocked by naloxone, an opiate antagonist — specifically implicating endogenous opiates in pain placebo effects, which would not be expected to benefit every possible outcome that might be measured.

3. When hearing that persons with a problem placed on a "placebo" report improvement, scientists commonly presume this must be due to the "placebo effect" - the effect of expectation/suggestion.

However, the effects are usually something else entirely. For instance: natural history of the disease, and regression to the mean. Consider a distribution, such as a bell-shape. Whether the outcome of interest is pain, blood pressure, cholesterol, or other, persons are classically selected for treatment if they are at one end of the distribution - say, the high end. But these outcomes are quantities that vary (for instance from physiological variation, natural history, measurement error...), and on average the high values will vary back down — a phenomenon termed "regression to the mean" that operates, placebo or no. (Hence, Hróbjartsson's findings.)

A different dece(i)bo problem beset Ted Kaptchuk's recent Harvard study in which researchers gave a "placebo," or nothing, to people afflicted with irritable bowel syndrome. They administered the placebo in a bottle boldly labeled "Placebo," and advised patients they were receiving placebos, which were known to be potent. The thesis was that one might harness the effects of expectation honestly, without deception, by telling subjects how powerful placebos in fact were - and by developing a close relationship with subjects. Researchers met repeatedly with subjects, gained subjects' appreciation for their concern and listening (as the researchers made clear), and repeatedly told subjects that placebos are powerful. Those placed on placebo obliged the researchers by telling them they had gotten better, moreso than those on nothing. The scientists attributed this to a placebo effect.

But what's to say patients weren't simply telling the scientists what they thought the scientists wished to hear? Such desire to please (a form, perhaps, of "social approval" reporting bias) had fertile grounds in which to operate and create what was interpreted as a placebo effect — which implies actual subjective benefit to symptoms. One wonders if so great an error of presumption would operate were there not an existing term, "placebo effect," to signify the interpretation the Harvard group chose.

Another explanation consistent with these results is specific physiological benefit. The study used a nonabsorbed fiber — microcrystalline cellulose — as the "Placebo" that subjects were told would be effective. The authors are applauded for disclosing its composition. But other nonabsorbed fibers benefit both constipation and diarrhea — symptoms of irritable bowel — and are prescribed for that purpose; psyllium is an example. Thus, specific physiological benefit of the "Placebo" to symptoms cannot be excluded.

Together these points illustrate that the term "placebo" cannot be presumed to imply "inert" (and generally does not); and that when studies see large benefit to symptoms in patients treated with "placebo" (expected from distribution considerations alone), one cannot infer these arose from large benefits of suggestion to symptoms (which evidence indicates may seldom operate).

Thus, rather than facilitating sound reasoning, evidence suggests that in many cases, including high stakes settings in which inferences may propagate to medical practice, substitution of a term — here, "placebo," "placebo effect" — for the concepts they are intended to convey, may actually thwart or bypass critical thinking about key issues, with implications to fundamental concerns for us all.