What is the placebo effect? And why trials need control groups

The placebo effect is an improvement in how someone feels that comes from receiving a treatment, and expecting it to help, rather than from anything active in the treatment itself. A placebo is the inactive version: a sugar pill, a saline injection, a sham procedure. The effect is real for some symptoms, mainly pain and nausea, but it is smaller and narrower than its reputation, and much of what looks like it is something else.

In short

• A placebo has no active ingredient; the placebo effect is the part of a patient's improvement that comes from the act of being treated and the expectations around it.
• Systematic reviews find placebos have little effect on objective outcomes, but they can ease patient-reported symptoms such as pain and nausea.
• That's why trials compare a treatment with a placebo, assigned at random and with nobody knowing who got what: it separates the drug's effect from everything else.

What is a placebo, and what is the placebo effect?

The word is Latin for "I shall please". A placebo is a treatment with no specific action on the condition: a tablet of sugar or chalk, an injection of salt water, or a sham procedure such as acupuncture needles that retract instead of entering the skin. The placebo effect is whatever change the placebo produces through the patient's expectations, conditioning and the experience of care.

The two are easily confused with a third idea, the placebo response: everything that happens to people in the placebo group of a trial. That response is usually much bigger than the true placebo effect, because people given a dummy pill get better, or seem to, for several reasons that have nothing to do with the pill.

Why the placebo group improves	What it is	Is it a placebo effect?
Natural course	Colds clear up, back pain eases and flare-ups settle on their own	No
Regression to the mean	People join trials when symptoms are at their worst, and extremes tend to be followed by more ordinary days	No
Reporting	Patients who want to please or feel they should improve may rate symptoms more kindly	No, though hard to tell apart
Other care	Trial patients are seen more often and may change habits	Partly
Expectation and conditioning	The brain changes how symptoms such as pain are felt	Yes

The second row matters most and is the least known. Regression to the mean alone can make an inert treatment look as though it works, because people tend to seek help, and enroll in studies, on their bad days.

How strong is the placebo effect?

Much weaker in general than once thought. An influential 1955 paper, The Powerful Placebo, established the idea that placebos often helped patients a great deal. A 1997 reanalysis of its data found it had not accounted for regression to the mean and offered no evidence of a placebo effect at all (Wikipedia).

The cleaner test is to compare a placebo with no treatment at all: the difference between those two groups is the placebo effect, with natural recovery and regression subtracted out. In 2001, the Danish researchers Asbjørn Hróbjartsson and Peter Gøtzsche analyzed 114 trials that did exactly that, in the New England Journal of Medicine. They found little evidence that placebos had powerful clinical effects. Placebos made no significant difference to objective outcomes or yes-or-no outcomes such as whether a patient recovered. They found possible small benefits for subjective, patient-reported outcomes, and for pain they estimated a reduction of 6.5 mm on a 100 mm pain scale, 6.5% of the scale's length.

Their larger Cochrane review in 2010, with outcome data from 202 trials across 60 conditions, reached a similar verdict: placebos had no important clinical effects in general, but in certain settings could influence patient-reported outcomes, especially pain and nausea. It added a caution: it is difficult to tell a real placebo effect on a reported symptom from biased reporting. The effect on pain varied a great deal between trials, from negligible to clinically important.

Still argued

How much the placebo effect matters in medicine is debated. Researchers such as Hróbjartsson and Gøtzsche read the trial evidence as showing small, mostly subjective effects that don't justify using placebos outside research. Others point to laboratory studies, going back to 1978, showing that placebo pain relief partly depends on the brain releasing its own opioids, and argue that the effect is a real, measurable process worth harnessing through how care is delivered. Both sides agree that placebos don't shrink tumors, cure infections or reverse the course of a disease, and that the effects that exist are largest for symptoms people report themselves.

Not medical advice

This note explains research on placebos. It isn't medical advice: decisions about treatment belong with a qualified clinician.

Why do trials use placebos and double blinding?

Because without them, nobody can tell how much of an improvement the treatment caused. Here is how a trial of a new pain drug is built, step by step, and what each step removes. The numbers are an illustration.

1. Recruit patients with the condition. Say 200 people with chronic knee pain, rated 60 on average on a 0 to 100 scale.
2. Assign them at random. A computer puts 100 in each group, so age, severity and every other factor are balanced by chance, which is the core of telling correlation from causation.
3. Give the control group an identical placebo. Same size, color, taste and schedule. Both groups now share natural recovery, regression to the mean, extra attention and expectation.
4. Blind the patients. If they knew which pill they had, expectation would differ between groups.
5. Blind the researchers too. Doctors who know who got the drug may treat, encourage or score patients differently. Blinding both sides is what "double-blind" means.
6. Compare the groups at the end. Suppose pain falls from 60 to 45 with placebo and from 60 to 35 with the drug.

The drug group improved by 25 points, but only 10 of them are the drug's doing. The other 15 points happened in both groups, so they belong to everything except the drug. A study with no placebo group would have credited the drug with all 25. Whether a 10-point gap is bigger than chance would produce is then a question for statistics, which a p-value helps answer.

Can a placebo work if you know it's a placebo?

Some small trials suggest it can, at least for symptoms people report themselves. In a 2010 study at Beth Israel Deaconess Medical Center in Boston, 80 people with irritable bowel syndrome were randomly assigned to no treatment or to pills they were told plainly were placebos, "like sugar pills", though they were also told that such pills had helped people with the condition in clinical studies. After three weeks, the open-label placebo group rated their overall improvement higher, 5.0 against 3.9 on a 7-point scale (PLOS ONE). The trial was small and its outcome was self-reported, so its authors called for further research rather than a change in practice.

What is the nocebo effect?

The nocebo effect is the reverse: harm or side effects that come from expecting harm. It shows up clearly in drug side effects. In the SAMSON trial, led from Imperial College London, researchers recruited 60 people between June 2016 and March 2019 who had stopped taking statins because of side effects. Each took a statin, an identical placebo or no tablet in randomly ordered months over a year, recording their symptoms on a phone app. According to the British Heart Foundation, which funded it, 90% of the symptoms people reported while taking the statin also appeared while they were taking the placebo.

Expectations shape how symptoms are felt, which ties both effects to how the brain builds experience from signals and predictions. And for anyone reading a health claim, the placebo is the first thing to rule out: an improvement without a comparison group is weak evidence that a treatment works.

Check yourself

• Without looking back, name three reasons people in a placebo group improve that aren't the placebo effect.
• In the knee-pain example, why is the drug's effect 10 points and not 25?
• What did comparing placebo with no treatment at all show about how strong the placebo effect is?

Sources

• Is the placebo powerless? An analysis of clinical trials comparing placebo with no treatment — Asbjørn Hróbjartsson and Peter Gøtzsche, New England Journal of Medicine, 2001
• Placebo interventions for all clinical conditions — Asbjørn Hróbjartsson and Peter Gøtzsche, Cochrane Database of Systematic Reviews, 2010
• Placebos without deception: a randomized controlled trial in irritable bowel syndrome — Ted Kaptchuk and colleagues, PLOS ONE, 2010
• The SAMSON trial — British Heart Foundation
• Placebo — Wikipedia
• Nocebo — Wikipedia