Why Placebo Response Is So Large in Depression Trials

The short answer

Placebo response is large in depression trials because the placebo arm captures far more than the effect of an inert pill. When a person in the inactive group improves, that change reflects the natural course of a mood episode, regression to the mean, the effect of repeated attentive contact with study staff, and the expectation that comes from knowing active medication might be in the capsule. Across many antidepressant trials, placebo response has hovered around 35 to 40 percent, a figure summarized in a 2020 analysis by Holper and Hengartner in BMC Psychiatry. Because that improvement is genuine and sizeable, it sits underneath the drug arm and shrinks the apparent distance between drug and placebo. This piece is about how to read that gap, not about how anyone should be treated.

What the placebo arm is actually measuring

It helps to separate two ideas that get blurred together. The placebo effect, in the narrow sense, is the improvement attributable to expectation and the ritual of treatment. The placebo response, the number a trialist actually observes, is everything that happens in the placebo group over the study period. Most of that observed response is not the mind healing itself. It is the sum of several ordinary phenomena.

Depression is episodic. Many episodes improve on their own, and people usually enroll in a trial when they feel worst, which is exactly when the next measurement is most likely to be better by chance alone. That statistical pull toward the average is regression to the mean, and it operates in every arm regardless of what pill is given. Layered on top is the natural history of the disorder, the tendency of a bad stretch to lift over weeks. Add the structured attention of a trial, frequent visits, rating scales, and clinicians who ask careful questions, and you have a setting that tends to make people feel and report better.

Why expectation is unusually potent here

Mood disorders are measured through what patients and raters say, not through a blood value or an imaging finding. Outcomes such as the Hamilton or Montgomery-Asberg scales are built from judgment. Subjective, rater-dependent endpoints are more responsive to expectation than a hard laboratory number would be, so the same expectation that nudges a headache trial can move a depression score substantially. When a participant knows there is a real chance of receiving an active antidepressant, that hope itself becomes part of the measured response.

Why the gap looks small even when drugs work

The most rigorous synthesis of antidepressant efficacy is the 2018 network meta-analysis by Cipriani and colleagues in the Lancet, which pooled 522 trials and roughly 116,000 participants across 21 drugs. Its central finding was that all of the studied antidepressants were more effective than placebo for acute major depression. Its second finding, easy to miss, was that the advantage was modest: odds ratios for response clustered in a low range, and the overall standardized difference from placebo was small.

Those two statements are not in tension. A drug can be reliably better than placebo and still separate from it by a narrow margin, because the placebo arm is not near zero. If placebo response captures 35 to 40 percent improvement, the drug has to climb above an already high floor. The measured gap is the drug response minus a large placebo response, so a big placebo number mathematically compresses the difference that appears in the results table. Reading the small effect size as evidence that the drugs do nothing is a misreading; it is evidence that the comparator is strong.

The rising-response story, and why it is contested

For years a tidy narrative held sway: placebo response had been climbing decade over decade, swallowing the drug-placebo difference and producing more so-called failed trials. Early reports described placebo response in published trials rising by several percentage points per decade across the 1980s and 1990s. Analyzing regulatory data from 1987 to 2013, Khan and colleagues reported in World Psychiatry that the magnitude of placebo response continued to grow, rising by roughly six percent since 2000, while the drug-placebo difference stayed about the same across the antidepressants approved in that window.

The story is less settled than it sounds. Furukawa and colleagues, reviewing 252 published and unpublished trials in Lancet Psychiatry in 2016, found that placebo response looked stable since about 1991 once they adjusted for how trials had changed: their length, the number of study centers, and the use of fixed dosing. In other words, part of the apparent rise was a byproduct of design drift rather than a genuine shift in patients. Baseline severity and certain design features moved the placebo number; calendar year, by itself, did much less than the headlines suggested.

Design features that push placebo response up

Several recurring factors are worth watching when you read a trial. More study sites tend to raise placebo response, partly through looser rating consistency across many raters. Enrolling patients whose baseline scores are inflated at intake leaves more room for regression to the mean. A higher probability of receiving active drug, as in a trial with several drug arms and one placebo arm, raises expectation in every participant. And unblinding matters: older drugs with obvious side effects could tip off raters, which, as Holper and Hengartner discuss, distorts the comparison in ways that can cut either direction.

How to read the gap

A large placebo response is not a flaw to be explained away; it is a signal that the outcome is soft, the disorder is self-limiting for many, and the trial setting is therapeutic in its own right. The practical lesson for interpretation is to resist two opposite errors. Do not treat a small drug-placebo difference as proof of no effect, and do not treat a strong placebo arm as proof the drug is unnecessary. Ask instead what the placebo arm was measuring, how sick the enrollees were at baseline, how many sites and arms the trial ran, and whether the blind likely held. This article is educational and is not medical advice.