Patients Told Their Pills Were Fake Still Got Better. 59% Reported Relief, and the Effect Has Held Up.

The Story Everyone Knows

Every medical student learns the same rule about placebos: they only work if the patient believes the pill is real, deception is the active ingredient, and a sugar pill handed over with a wink is a joke, not a treatment, so any doctor who prescribed one honestly would be prescribing nothing at all.

That rule felt obvious because it rested on a clean theory: the placebo effect is expectation, expectation requires belief, and telling someone the pill is inert should collapse both at once, so for decades nobody tested it.

The Study

Ted Kaptchuk, a Harvard Medical School researcher who had spent his career studying placebo effects, decided to run the test anyway, recruiting 80 patients with irritable bowel syndrome, a chronic gut disorder, diagnosed by Rome III criteria.

Patients were randomized to two groups for three weeks. One group received open-label placebo pills, presented with a careful script about "placebo pills made of an inert substance, like sugar pills, that have been shown in clinical studies to produce significant improvement in IBS symptoms through mind-body self-healing processes." The other group received no treatment at all but got the same quality of interaction with providers, and patients were told they need not even believe, only take the pills twice a day.

At the 21-day endpoint, the placebo group scored 5.0 on the 7-point Global Improvement Scale against 3.9 for the no-treatment group (p = .002), a gap already visible at the 11-day midpoint (5.2 versus 4.0, p < .001), and on the trial's practical measure, adequate relief, 59 percent of placebo patients reported feeling better against 35 percent of controls, with severity and quality of life trending the same way.

Senior author Anthony Lembo admitted he had not expected it to work. The trial was small, single-center, and short.

The Arithmetic Nobody Did

Here is the calculation the coverage skipped: adequate relief ran 59 percent against 35 percent, an absolute gap of 24 percentage points, and dividing 100 by 24 gives a number needed to treat of about 4, meaning about one extra patient in four reports adequate relief.

That matches many approved IBS treatments, which often help one in five or one in eight patients beyond placebo, except here the "treatment" has no active ingredient, no side effects, and nothing to overdose on.

The standardized effect works out to roughly 0.78 standard deviations, a large effect by the usual benchmarks, computed from the endpoint means of 5.0 and 3.9 with standard deviations of 1.5 and 1.3.

The pattern did not stay a one-off: in 2016, Carvalho and colleagues randomized 97 chronic low back pain patients to open-label placebo plus usual treatment or usual treatment alone, and after three weeks pain fell 1.5 points on a 10-point scale in the placebo group against 0.2 in controls, disability fell 2.9 points against zero, and the switched control group improved by the same amount.

A 2021 meta-analysis by von Wernsdorff and colleagues pooled 11 open-label placebo trials across back pain, cancer fatigue, depression, and IBS, finding a standardized mean difference of 0.72, and a later, larger meta-analysis put the clinical-sample effect at 0.47, smaller but still solidly positive, so honest placebos kept working even as the effect size shrank toward something more believable.

The Strongest Counterargument

Now the case against, stated at full strength: every open-label placebo trial shares a fatal design flaw, since nobody can be blinded. Patients know they are getting the special treatment, researchers know who got it, and both sides have reason to report improvement, especially after patients were told, repeatedly and warmly, that placebos are powerful, so they may simply have told the researchers what the researchers wanted to hear.

That is not speculation: the 2010 Cochrane review by Hróbjartsson and Gøtzsche pooled 202 placebo-versus-no-treatment trials and found small effects overall, a relative risk of 0.93 for binary outcomes and a standardized mean difference of just 0.23 for continuous ones, with no evidence of important clinical effects in general, and much of what looks like a placebo effect is regression to the mean, since people enroll in trials when symptoms flare and flares fade on their own.

The sharpest cut comes from the newest data: a 2023 meta-analysis of open-label placebo experiments in healthy volunteers found a real effect on self-reported outcomes (SMD 0.43) and essentially zero on objective ones (SMD -0.02), so where machines measure blood pressure, wound healing, or grip strength, the honest placebo does nothing, and the effect lives in what people say they feel, which is exactly where reporting bias lives too.

Durability failed too: a 3-year follow-up of the back pain trial found no long-term benefit. The relief did not last.

Put bluntly: open-label placebos reliably change how people report symptoms for a few weeks, and nobody has shown they change the underlying disease.

What We Didn't Prove

This evidence does not prove that honest placebos treat disease: IBS and chronic back pain are symptom-driven conditions with no blood test to contradict a patient's report, and the original trial enrolled mostly women (70 percent) at a single academic center, so the result may not travel.

It does not prove the mechanism: conditioning, expectation, the ritual of taking pills, and the warmth of the clinical encounter are all tangled together in these trials and no study has cleanly separated them, and it does not prove the effect lasts beyond three weeks without the persuasive script and supportive clinician, or works for an objective marker like blood sugar.

And it does not prove that internet sugar pills are medicine: the trials used a specific script, a real clinician relationship, and fully informed patients.

The Bottom Line

What survives contact with the critics is narrower than the headlines: a pill with nothing in it, prescribed honestly, repeatedly beats doing nothing for subjective symptoms, across conditions, with an effect size that shrinks under scrutiny but does not vanish, and the most plausible reading is that the ritual of treatment itself, the pill, the schedule, the clinician who explains it, is doing real work on how symptoms are experienced even when everyone in the room knows the pill is inert.

That does not make deception obsolete or honesty magic: it makes the clinical encounter a measurable ingredient, one medicine has dosed blindly for a century.

What You Can Do

If IBS or chronic back pain treatments leave you short, ask your clinician about open-label placebo as an add-on, not a replacement. Cite Kaptchuk 2010 and Carvalho 2016.

Never stop a prescribed drug to try a placebo. No trial tested swapping a real drug for sugar pills.

If you try it, treat it as an experiment with a deadline. Track one symptom score for three weeks and stop if nothing changes.

Beware of sellers. Honest placebo in the trials came with a clinician, a script, and full disclosure; a bottle marketed online with vague healing claims is not the same thing.

Clinicians: the cheapest intervention in your toolkit may be the explanation itself. A warm, specific rationale about treatment rituals is part of what the trials delivered.