The One Rule Everyone Agrees On
If you ask a hundred doctors to name a single fact about placebos, most will land on the same answer: they only work if the patient doesn't know. The sugar pill heals because the patient believes it's medicine, and removing that belief should extinguish the effect entirely. This principle is taught in medical schools, baked into the architecture of clinical trial design, and treated as so obviously true that it barely registers as a claim requiring evidence of its own.
A growing body of research suggests this assumption is probably wrong.
The First Sugar Pill That Announced Itself
In 2010, Ted Kaptchuk's research group at Harvard Medical School ran an experiment that sounded absurd. Give eighty IBS patients pills in a bottle labeled "placebo," tell them the pills contain no medication, and see if they improve. Eighty patients with irritable bowel syndrome were randomized into two groups, with one group receiving the labeled placebo pills and the other receiving nothing beyond routine care, with both groups receiving the same quality of interaction with their providers (DOI: 10.1371/journal.pone.0015591).
The patients improved. Not slightly. On the IBS Global Improvement Scale, placebo patients scored 5.0 versus 3.9 at the 21-day endpoint (p = .002), with corresponding reductions in symptom severity and increases in adequate relief that were statistically significant across every measure the researchers tracked.
The bottle said "placebo." The providers said "placebo." And the patients reported feeling meaningfully better.
Sixty Trials Later, the Finding Holds
What looked like a one-off curiosity in 2010 has become one of the more robust findings in modern placebo research. In a 2025 meta-analysis published in Scientific Reports, Johannes Fendel and colleagues at the University of Freiburg synthesized 60 randomized controlled trials involving 4,554 participants across seven medical conditions and multiple experimental paradigms (DOI: 10.1038/s41598-025-14895-z).
The pooled effect was a standardized mean difference of 0.35 (95% CI: 0.26–0.44, p < 0.0001), which qualifies as small by Cohen's conventions but masks an important split between populations. Patients with actual medical diagnoses showed an SMD of 0.47, approaching medium territory, while healthy volunteers in laboratory experiments responded at a weaker 0.29. The sickest participants responded most to pills they knew were empty.
That SMD of 0.47 translates to a probability of superiority of approximately 63%: if you randomly selected one patient from the open-label placebo group and one from the control group, there is a 63% chance the placebo patient reports less severe symptoms. A coin flip gives you 50%, so a labeled sugar pill shifts the odds 13 percentage points without any active ingredient entering the bloodstream.
The meta-analysis tested whether the persuasive rationale typically accompanying open-label placebos was driving the results, and it was not; nor did the type of control group moderate the effect, whether the comparison was to no treatment, treatment as usual, or waiting list controls. These results survived sensitivity analyses excluding outliers, high-risk-of-bias studies, and publication bias adjustments, and the study was pre-registered on PROSPERO with full PRISMA 2020 compliance.
But Here's What Should Give You Pause
The strongest criticism of open-label placebo research comes not from people who doubt the data, but from those who question what "open-label" actually means in practice.
David Gorski, writing at Science-Based Medicine, has argued that the standard Kaptchuk protocol involves a 15-minute script telling patients that the placebo effect is powerful, that their bodies can respond automatically like Pavlov's dogs salivating at a bell, that a positive attitude helps but isn't necessary, and that taking the pills faithfully is critical. That is not disclosure. That is persuasion. Patients are told the pills are inert, but they are simultaneously told by a clinician in a medical setting that inert pills have been scientifically demonstrated to produce real healing.
This critique carries genuine weight, because the distinction between "honestly prescribing a placebo" and "persuading someone that a known placebo will work" is philosophically meaningful, and it remains unclear whether the effect would survive if a clinician simply handed a patient a bottle labeled "sugar pills" and said nothing else. If you stripped away the rationale entirely, would anything happen? Fendel et al. found that suggestiveness did not moderate the overall effect, but only five trials used a low-suggestiveness rationale, and the confidence intervals were wide enough that the question remains unresolved.
Then there is the matter of objectivity, which may be the most important caveat of all. When the researchers separated self-reported outcomes from objective physiological measures, the picture shifted sharply: self-report outcomes showed a clear effect (SMD = 0.39), while objective outcomes showed almost nothing (SMD = 0.09, not statistically significant). This echoes Hróbjartsson and Gøtzsche's landmark Cochrane reviews in 2001 and 2010, which found that placebos have no clinically meaningful effects on objective or binary outcomes. Open-label placebos change how people report feeling, but they do not shrink tumors, clear infections, or normalize blood chemistry.
What We Didn't Prove
Most of the 60 included studies were small, with sample sizes ranging from 9 to 133 participants. The participant pool skewed 72% female with a mean age of 30.5 years, and no individual study enrolled more than 133 people, so a genuinely definitive trial would require hundreds of participants with adequate statistical power.
Heterogeneity was moderate (I² = 53%), meaning the true effect likely varies across conditions in ways not captured by a single pooled estimate. The prediction interval was broad enough that in some populations the true effect may approach zero, and 20% of included trials carried high risk of bias.
Nearly all positive effects were confined to self-reported outcomes. This matters. If you have a bacterial infection, cancer, or diabetes, a sugar pill you know is a sugar pill will not alter the course of your disease.
The Bottom Line
The foundational assumption that placebos require deception is probably incomplete. Across 60 RCTs and over 4,500 participants, honestly labeled placebos produced small but statistically robust improvements in subjective symptoms like pain, fatigue, and digestive distress, with the strongest effects appearing in clinical populations with genuine suffering.
The mechanism likely involves expectation, ritual, and the therapeutic encounter rather than pharmacology. Placebos are not magic and they cannot cure disease, but they appear to be a clinically exploitable feature of how human brains process the experience of being treated, and unlike every other treatment in medicine, they carry essentially no risk of adverse effects.
What You Can Do
If you live with chronic pain, IBS, or cancer-related fatigue: Ask your doctor whether an open-label placebo approach might complement your current treatment. Several clinical trials are actively enrolling, and the evidence is strongest for conditions dominated by subjective symptoms.
If you're a clinician: The ethical barrier that historically prevented honest use of placebos may no longer apply. Kaptchuk's 2018 BMJ analysis argued for cautious clinical exploration, and the AMA's Council on Ethical and Judicial Affairs permits placebo use with patient consent when the goal is symptom management.
If you take supplements with weak evidence: Consider whether the benefit you experience might be a placebo effect operating through ritual and expectation, and whether a labeled sugar pill might achieve the same result at lower cost with fewer interactions.
For everyone: Be skeptical of anyone extrapolating these findings to claim placebos can cure serious diseases. The effect is small, concentrated in subjective outcomes, and absent from every objective biomarker measured across 60 trials.