Placebo Effect

← Encyclopedia

People given a pill that contains nothing — sugar, starch, a saline drip — sometimes report that they feel better. Not because they are lying, and not always because they would have recovered anyway. The puzzle is how much of that improvement is real, what kind of “better” it is, and whether anything has actually changed in the body or only in the report of it. The placebo effect is the name medicine gave to that puzzle, and the history of the name is largely the history of getting the answer wrong, then correcting it.

“Placebo” is Latin for I shall please — the future tense of to please — and the verb tells the truth about how the word began. It entered medicine in the late eighteenth century for a remedy given more to satisfy a patient than to treat him; by 1811 a medical dictionary defined it as “any medicine adapted more to please than to benefit the patient.” The phrase carried a faint contempt, and clinicians of the period already knew the trick from the inside. Asked by a patient whether to take a fashionable elm-bark panacea, the Paris physician Bouvart is supposed to have answered: take it, and hurry, while it still cures. Its dark twin came much later — nocebo, “I shall harm,” coined in 1961 for the symptoms that negative expectation can produce on its own.

A distinction has to be set down before anything else makes sense, because almost all the confusion in the subject comes from skipping it. In a trial, everything that improves in the group given a dummy treatment is the placebo response — and most of that has nothing to do with the placebo. The placebo effect proper is only the part that remains after subtracting what would have happened with no treatment at all. People recover on their own; symptoms wax and wane; the very sick, measured at their worst, tend to test better the next time by simple statistics. When the two are conflated, the placebo looks miraculous.

That is exactly the mistake the field’s founding paper made. In 1955 Henry Beecher published “The Powerful Placebo,” pooled fifteen studies, and reported that about a third of more than a thousand patients — the figure he gave was thirty-five per cent — had been “satisfactorily relieved by a placebo alone.” The paper fixed the idea in medicine’s mind that placebos were clinically potent, a standing demonstration of the brain’s hold over the body. It was cited for decades. It was also wrong in the way that matters most.

The audit came in two waves. In 1997 Kienle and Kiene went back to the very studies Beecher had cited and found, in their reading, no genuine placebo effect in any of them — the improvements were accounted for by spontaneous recovery, fluctuation of symptoms, regression to the mean, parallel treatment, and a list of softer biases: answers of politeness, mis-scaling, misquotation. The placebo topic, they wrote drily, “seems to invite sloppy methodological thinking.” Then in 2001 Hróbjartsson and Gøtzsche did the thing Beecher had not: they pooled only trials that included both a placebo arm and a no-treatment arm, the single design that can isolate the effect from everything else. Across that evidence, and again in their larger Cochrane review of 2010 — over two hundred trials, sixty conditions — they found little sign that placebos have important effects on disease in general. The powerful placebo, measured properly, mostly evaporated.

But not entirely, and the residue is the interesting part. Even the deflationary reviews found a real, if modest, signal in one place: patient-reported outcomes, above all pain and nausea — the felt symptoms, the things only the patient can measure. And there the effect is not merely a reporting artefact, because it has a chemistry. In 1978 Levine, Gordon and Fields gave naloxone — a drug that blocks opioid receptors — to people whose dental-surgery pain had been eased by a placebo, and the relief reversed. If switching off the body’s own opioids switches off the placebo’s analgesia, then that analgesia was an opioid event, a real physiological cascade set off by an inert pill. Later work filled in the picture: placebo responses run on expectation and on conditioning, on the meaning and ritual of the therapeutic act, and they recruit identifiable circuitry — opioids and dopamine, the prefrontal cortex weighing what the situation has taught the body to predict. Benedetti, who has mapped much of this, frames placebo effects as the body’s response to the whole context of treatment, and puts the thesis bluntly: words and rituals can modulate the same biochemical pathways that drugs do. There is not one placebo effect but many, best charted in pain and in Parkinson’s disease.

Here is the line that holds the whole subject together, and it must be stated plainly: the placebo moves suffering, not pathology. It can lower the pain a person feels and the nausea they report; it does not shrink a tumour or clear an infection or alter the course of the disease underneath. The 2010 review put it in those terms — placebos did not appear to affect the diseases themselves, only outcomes that pass through a patient’s perception. That single boundary is why documenting the effect is not the same as recommending it, and why the honest version of the story is so much smaller than Beecher’s.

Two further findings complicate it without crossing that line. Patients told outright that their pills contained no active medicine — inert sugar pills, described as exactly that — have still improved in trials of irritable bowel syndrome, which suggests the effect does not strictly require deception; a 2017 meta-analysis pooled five such studies and found a real signal, while warning that the trials were few, unblinded, and bundled the honest pill with an encouraging script that cannot be separated from it. And the nocebo runs the same machinery in reverse: warned of side effects, patients reliably report them, sometimes from inert pills, sometimes from no pill at all. Expectation cuts both ways. It is precisely because expectation moves results that medicine learned to defend against it — to give half the subjects a sham, conceal which is which, and blind the researchers too. The defence has a long pedigree: in 1799 John Haygarth tested fashionable metal healing-rods against identical rods carved from plain wood, and found the wood worked just as well.

That last experiment points back across the whole history of healing. Long before there were trials, there were shrines, rituals, the laying-on of hands, charismatic healers, and remedies now known to be inert — and a great many of them, by every account, made people feel better. Kaptchuk, who has studied the placebo as the performance of a healing ritual rather than the action of a dummy pill, argues that what does the work is the broad weave of the encounter: attention, the practitioner’s confidence, the patient’s expectation, the drama of the setting. His group tried to take that apart in a study of irritable bowel syndrome, escalating the treatment from bare observation, to sham acupuncture, to sham acupuncture delivered by a warm and attentive clinician — and watched the rate of relief climb from twenty-eight to forty-four to sixty-two per cent, with the relationship itself the strongest ingredient. The placebo effect, on this reading, is the laboratory’s name for what ritual healing always traded in: the measurable power of meaning and expectation over the experience of being ill.

The kinship holds only with its limits attached. The mechanism is genuine — it has receptors and circuits and a name — and it is bounded, and the boundedness is the whole point. It is not magic, and it does not heal the body’s hidden damage; it works on the part of illness a person can feel, by way of the meaning they assign to the care they receive. Medicine spent a century first overstating that power and then measuring it down to its true size. What is left, once the arithmetic is done, is smaller than the old word promised and stranger than the debunking allowed: an inert thing that does nothing, given inside a relationship that does something.

→ Related: Mesmerism