How Researchers Estimate Antidepressant Discontinuation Symptoms

The short answer

Published estimates of how often people develop discontinuation symptoms after stopping an antidepressant range from roughly 15% to well over 40%, and most of that gap is an artifact of how each study was built rather than a genuine disagreement about biology. The single most consequential design choice is whether a review subtracts the symptoms that also appear when people stop placebo; apply that correction and lean on trials that measured symptoms in a defined window, and roughly one in six to seven people report discontinuation symptoms. Omit placebo comparison, weight long-term users, and pool observational surveys, and the figure climbs. Reading any single headline number without asking how it was assembled will mislead you.

Why the estimates split

In 2024, Henssler and colleagues published a systematic review and meta-analysis in The Lancet Psychiatry pooling 79 studies and just over 21,000 participants, about 16,500 of whom stopped an antidepressant and about 4,500 of whom stopped placebo. They found that around one in three people who stopped an antidepressant reported at least one discontinuation symptom, but so did roughly one in six who stopped placebo. Subtracting that placebo-associated background left an incidence attributable to the drug itself of roughly 15%, or one in six to seven patients, with severe symptoms in under 3%, or about one in 35.

A separate 2024 meta-analysis in Molecular Psychiatry by Zhang and colleagues, pooling 35 studies, reported a much higher pooled incidence of antidepressant withdrawal symptoms, around 43% overall and 44% across randomized trials. Both papers are peer-reviewed and technically competent. They diverge because they asked slightly different questions of a different slice of the literature and handled the control group differently.

The placebo problem

Stopping an inert pill also produces headaches, dizziness, irritability, and sleep disruption, driven by expectation, by the natural return of the underlying condition, and by the ordinary noise of everyday symptoms. This is the nocebo effect. A review that reports the raw proportion of people with any symptom after stopping the drug captures all of that background. A review that subtracts the placebo rate tries to isolate what the medication specifically caused. Neither is wrong, but they answer different questions: what will a patient experience versus what does the drug cause. Estimates that skip placebo correction will almost always look larger.

Duration of use

Most randomized trials that measured discontinuation were designed to test whether an antidepressant works, not what happens when people come off after years of treatment. As a result, exposure in the pooled trials is short. A 2025 appraisal and reanalysis in Psychological Medicine calculated that the weighted average duration of antidepressant exposure across the trials in the Henssler dataset was under six months, about 23 weeks, and that in most of those studies participants had taken the drug for less than three months. That poorly matches the real world, where the same appraisal notes about half of people on antidepressants in the UK have taken them longer than a year, and a similar share in the US for more than five years. If duration of exposure raises the likelihood or intensity of withdrawal, short-exposure trials will understate what long-term users face. This is a genuine limitation of the trial-heavy estimate, not a flaw a reviewer can simply correct away.

How symptoms were counted

Numbers also move depending on the measuring instrument. Structured checklists such as the Discontinuation-Emergent Signs and Symptoms scale ask about many symptoms and will flag more people than spontaneous reporting, where a patient has to volunteer a complaint unprompted. The follow-up window matters too. Several trials assessed symptoms only for about two weeks after stopping, which can miss slower-onset or longer-lasting courses. A review built on sensitive checklists and long windows will report higher incidence than one built on brief spontaneous reporting, before any biology enters the picture.

Which studies get included

The most dramatic divergences come from study selection. When the Psychological Medicine group re-analyzed a subset of the same evidence, it reached a higher figure partly by narrowing to five studies with 601 participants and setting the placebo comparison aside. Reasonable researchers can defend restricting a sample to the most relevant populations, but each inclusion rule shifts the result, and a reader cannot judge a number without seeing those rules. This is why the honest summary is a range with stated assumptions, not a single definitive percentage.

What the tapering evidence does and does not show

There is broad agreement that stopping abruptly is worse than stopping gradually, and that some drugs with short half-lives provoke more symptoms than others. The proposed remedy is hyperbolic tapering, which reduces the dose in progressively smaller steps because the relationship between dose and serotonin-transporter occupancy is curved rather than linear, so equal dose cuts do not produce equal biological changes near the bottom of the range.

The mechanistic rationale is coherent, and observational data are encouraging. A prospective Dutch cohort using tapering strips reported that withdrawal during slow, daily-step tapering was limited and inversely related to how fast the dose came down, with a short-term success rate around 71%. What is thinner is high-quality randomized evidence directly comparing hyperbolic tapering against conventional tapering for both symptom burden and successful, durable discontinuation. Encouraging cohorts are not the same as randomized proof, and cohorts of motivated patients using a structured program can look better than routine care for reasons unrelated to the taper shape. The reasonable reading is that gradual, individualized tapering is sensible and mechanistically justified, while the precise schedule that best balances comfort against relapse is still being established.

How to read the next headline

When you see a withdrawal-incidence figure, ask five questions. Was it placebo-corrected? How long had those patients taken the drug? How were symptoms measured, and for how long after stopping? Which studies were included or excluded? And is the outcome any symptom, or clinically significant or severe symptoms? The difference between 15% and 43% lives almost entirely in those answers. The signal underneath is consistent and worth stating plainly: discontinuation symptoms are real, most are mild, a minority are severe and disabling, and slower tapering generally helps.

This article is educational and not medical advice; decisions about starting, changing, or stopping any medication belong with a qualified clinician who knows the individual.