How common are antidepressant withdrawal symptoms? The honest answer is: it's contested

How often stopping an antidepressant causes withdrawal symptoms is one of the genuinely unsettled questions in psychiatry, and the honest answer is a range, not a number. The largest meta-analysis puts the incidence of symptoms attributable to the drug itself at roughly 15% — about one in six or seven people — while critics who reanalysed a subset of the same studies put it above 50% [s1][s2].

The disagreement is not a rounding error. It turns on which studies you trust to measure withdrawal at all, and that makes it a useful case in how a single clinical fact can be reported three different ways depending on the evidence you admit.

What the largest meta-analysis found

The 2024 synthesis in The Lancet Psychiatry pooled 79 studies — 44 randomised controlled trials and 35 observational studies — covering 21,002 patients (72% female, mean age 45) who stopped either an antidepressant or placebo [s1]. The incidence of at least one discontinuation symptom was 0.31 (95% CI 0.27 to 0.35) after stopping an antidepressant, and 0.17 (95% CI 0.14 to 0.21) after stopping placebo [s1].

That placebo figure is the crux. If nearly one in five people report withdrawal-type symptoms after stopping a dummy pill, then a large share of what looks like drug withdrawal is a non-specific or nocebo effect — the expectation of symptoms producing them [s1]. Restricting to the randomised trials, where drug and placebo can be compared directly, the summary difference in incidence was 0.08 (95% CI 0.04 to 0.12) [s1]. Folding that placebo rate in, the authors conclude the incidence specifically attributable to the antidepressant is approximately 15%, affecting one in six to seven patients [s1].

Severe symptoms were less common: an incidence of 0.028 (95% CI 0.014 to 0.057) after an antidepressant against 0.006 (95% CI 0.002 to 0.013) after placebo [s1]. Not every drug behaved the same — desvenlafaxine, venlafaxine, imipramine and escitalopram were associated with higher frequencies of symptoms, and imipramine, paroxetine and desvenlafaxine or venlafaxine with greater severity [s1]. Heterogeneity across studies was, in the authors' own word, substantial [s1].

Why critics say that understates it

A 2025 reanalysis in Psychological Medicine accepts almost none of that framing [s2]. Its authors argue that most of the data in the larger review come from pharmaceutical-industry efficacy studies in which withdrawal was a minor, spontaneously-reported consideration rather than something systematically measured, and that prior antidepressant use in those studies was short — 12 weeks or less in all but one of the trials they retained [s2]. Spontaneously reported adverse events, they argue, undercount a syndrome nobody was formally screening for.

When they repeated the meta-analysis using only the five studies that assessed withdrawal with a systematic, relevant method, the pooled incidence of any withdrawal symptom was 55% (95% CI 31% to 81%; N = 601), without subtracting a nocebo effect, and with high heterogeneity [s2]. Their conclusion is blunt: the larger review is "based on unreliable data" and does not provide an adequate basis for evaluating withdrawal [s2].

Both things can be read off the page at once. A reader is not choosing between a careful study and a sloppy one; they are watching two competent teams disagree about whether industry efficacy trials can measure a harm those trials were not built to detect — and about whether a high placebo rate is a real signal to subtract or an artefact of the same weak measurement.

The estimate that predates the fight

The dispute is older than either paper. A 2019 systematic review in Addictive Behaviors — influential in pushing guidelines toward slower tapering — found withdrawal incidence across 14 studies ranging from 27% to 86%, with a weighted average of 56% [s3]. Of four studies rating severity, a weighted average of 46% of those affected endorsed the most extreme severity rating offered [s3]. And seven of ten studies on duration contradicted then-current UK and US guidance that framed withdrawal as brief, finding a meaningful share of people had symptoms for more than two weeks, sometimes months [s3].

So the field's three reference points are a pre-2020 review near 56%, a 2024 meta-analysis netting about 15% once placebo is accounted for, and a 2025 reanalysis back near 55% [s1][s2][s3]. The clinical direction of travel — toward gradual, often hyperbolic tapering rather than abrupt cessation — has held across all three, because it is the cautious response whether the true rate is one in six or one in two [s3].

What is actually settled

That discontinuation symptoms are real, that some drugs and shorter half-lives carry more of them, and that a non-trivial number of people have a genuinely difficult time stopping [s1][s3]. What is not settled is the headline incidence, and any single confident percentage — in either direction — is overstating the certainty the evidence supports.

This is a question for a prescriber, not a number to act on alone: the coverage here describes a scientific dispute and is not medical advice, and nothing in it should be used to start, stop, change or taper a medication. For the adjacent and better-characterised problem of stopping sedatives, see the benzodiazepine tapering guidance; for how antidepressants compare with talking therapy in the first place, see therapy versus antidepressants.

Sources

• Incidence of antidepressant discontinuation symptoms: a systematic review and meta-analysis — The Lancet Psychiatry, 2024-06-05
• Evidence on antidepressant withdrawal: an appraisal and reanalysis of a recent systematic review — Psychological Medicine, 2025-07-22
• A systematic review into the incidence, severity and duration of antidepressant withdrawal effects: Are guidelines evidence-based? — Addictive Behaviors, 2018-09-04