Stopping antidepressants: what the evidence says about withdrawal and relapse

"How do I come off this?" is one of the most-asked questions about antidepressants and one of the worst-served by the internet, because two completely different things happen when people stop and the two get blurred together constantly.

Withdrawal is your body reacting to the drug leaving. Relapse is the depression coming back. They have different timelines, different symptoms, and different answers — and the evidence base for each is a different shape.

Here is what the studies in our library actually found, with every claim linked to the study summary behind it.

This is information to bring to your prescriber, not medical advice, and not a reason to change anything on your own. Stopping an antidepressant abruptly is specifically what the evidence here argues against.

What the biggest stopping trial found

Most of what gets said about relapse rates comes from trials that were not designed to answer the question. One was.

ANTLER randomised 478 primary-care patients who already felt well enough to stop long-term antidepressants — either to taper off, or to stay on. Over the following year, 56% of those who stopped relapsed, versus 39% of those who continued [1].

Read that in both directions, because most coverage only reads it in one:

• Stopping roughly increased the chance of relapse within a year from about two in five to about half.
• And 44% of the people who stopped did not relapse in that year. Stopping is not futile. It is a real risk with real odds, not a guaranteed return to where you started.

The trial cannot tell you which group you are in. Nobody can, yet. What it can tell you is that the decision has a measurable cost and that the cost is not certainty.

Study summary: Maintenance or Discontinuation of Antidepressants in Primary Care — gold standard, 87/100

One honest note about our own library: that paper is indexed twice, once under sertraline and once under fluoxetine, because the trial covered both. It is one study, not two, and it would be wrong to count it as two pieces of evidence. You can see what each drug is actually approved for on the sertraline and fluoxetine approval pages.

Withdrawal is a different question, with weaker evidence behind it

Relapse got a large randomised trial. Withdrawal did not.

The best synthesis in our library reviewed 61 reports, including 22 double-blind randomised trials, and found withdrawal symptoms after stopping any SNRI — occurring most often with venlafaxine, typically starting within days and lasting weeks, and appearing even with gradual tapering [2].

That last clause matters. Tapering helps; the review found it does not reliably prevent the problem.

Study summary: Withdrawal Symptoms after SNRI Discontinuation — early signal, 52/100

We are showing you a weak score on purpose. That review scores 52 out of 100 in our library, and the caveat recorded against it says why: it pools randomised discontinuation phases, open-label trials and case reports together, so withdrawal incidence cannot be precisely quantified from it. The design was scored conservatively because it includes observational evidence.

So the honest position is: withdrawal after stopping an SNRI is well documented and real, and how often it happens is not established. Anyone quoting you a confident percentage — high or low — is going beyond what this evidence supports. If you are on venlafaxine specifically, it is the drug the review flagged most often.

The label change most people have never heard about

On 16 May 2019, the European Medicines Agency's safety committee ordered every manufacturer of citalopram, escitalopram, fluvoxamine, fluoxetine, paroxetine, sertraline, duloxetine, venlafaxine, desvenlafaxine and milnacipran to add this warning within two months [3]:

"There have been reports of long-lasting sexual dysfunction where the symptoms have continued despite discontinuation of SSRIs/SNRI."

The patient-leaflet wording is gentler: "In some cases, these symptoms have continued after stopping treatment."

Study summary: PRAC signal recommendation, May 2019 — strong evidence, 70/100

Be careful with this one in both directions, and our record says so explicitly. A regulatory label change is a precautionary act on a safety signal, not proof of incidence, causation or permanence. What the committee weighed was pharmacovigilance reports, published literature, social media and manufacturer reviews — not a prevalence study. How often this happens is unknown.

That is genuinely different from "it is rare" and genuinely different from "it is common". It is unmeasured, and a regulator decided the signal was strong enough that people should be told before they start. If you want the wider picture on this, we have a separate page on SSRIs and sexual side effects.

Why the trials cannot tell you how long is long enough

There is a structural gap under all of this. When we counted the trial length of every randomised trial in our library, 47 of the 57 that state a duration measured their headline result at three months or sooner — while people routinely stay on antidepressants for years.

ANTLER is one of the rare exceptions, following people for a full year. Most of the evidence about these drugs does not run long enough to answer the questions people actually have about staying on them or coming off them.

The full trial-duration audit is here, with every trial listed and its horizon quoted from the record.

Questions worth taking to your prescriber

Each of these exists because of a specific finding above:

• If I stop, what taper schedule would you use, and over how long? (Tapering helps; the SNRI review found it does not reliably prevent withdrawal, so the plan matters.)
• How would we tell withdrawal apart from relapse if I start feeling bad? (Timing and symptom type are the usual clues — this is the distinction the whole page rests on.)
• What is the check-in plan for the first three months after stopping? (ANTLER's relapses accumulated over a year, not a week.)
• Am I on an SNRI, and does that change the plan? (Withdrawal was reported most often with venlafaxine.)
• Given the 2019 EU label change, what should I know about sexual side effects before starting or stopping?
• What would make you say "now is not the right time to stop"?

Why this page shows its working

Every study summary linked here is free to read on the web with no account, and each carries a trust score calculated from recorded facts about the paper — study design, funding, journal, sample size, preregistration — rather than typed in by an editor. Where a study has a specific weakness, a written caveat sits next to the score rather than beneath it. That is why the withdrawal review above shows a 52 instead of being quietly presented as settled.

The scoring rubric is public, including the parts that do not flatter us. Browse the whole evidence library if you want to check our work, or read who funded the research behind your medication.

If you are in crisis right now, please do not wait: call or text 988 (Suicide & Crisis Lifeline) or text HOME to 741741 (Crisis Text Line). Both are free, confidential and staffed 24/7.