Coming off antidepressants

Most people can stop an antidepressant safely. Some have a hard time, and a few have a very hard time. Here's what the evidence says about who gets withdrawal symptoms, how to tell them from relapse, and how to taper in a way that gives you the best odds, for every common antidepressant.

By Reid Robison, MD. 3,892 words, with a medicine-by-medicine explorer, a withdrawal-or-relapse checker and a taper planner to bring to your prescriber.

Researched and reviewed by Reid Robison, MD × OpenEvidence × Claude How this post was made

This question came up in my clinic today, about one medicine. A patient had tried twice to come off venlafaxine and both times felt dizzy, sick and frightened within three days. She wasn't relapsing, and she wasn't doing anything wrong. She was coming down a staircase that had been built with the steps the wrong size. This post is the general version of the conversation we had.

1Withdrawal, discontinuation, dependence

You'll see two names for the same thing. "Discontinuation syndrome" is the term in the DSM and on drug labels.⁴ "Withdrawal" is what patients call it, and it's the term UK guidance now uses.⁹ I use both. What matters is what the words don't mean.

They don't mean addiction. Addiction involves craving, loss of control and use despite harm. Antidepressants don't do that. What they can do is produce physical dependence: the brain adapts to a medicine that's been there for months or years, and it needs time to adapt back when the medicine leaves. Blood pressure medicines and steroids do the same. It's a property of the body, not a flaw in the person.

2What it feels like

Clinicians remember the common symptoms with the word FINISH.¹⁰ Tap a letter.

Timing is the other clue. Symptoms usually start two to four days after a dose is cut or stopped, peak within the first week, and settle within about two weeks.⁴¹¹ The shorter a medicine's half-life, the sooner and sharper the start. Fluoxetine leaves the body so slowly that symptoms, if they come at all, can show up weeks later.¹²

For a minority, symptoms last much longer than two weeks. In one survey of long-term users, 30 percent of those who had symptoms said they lasted more than three months.³ The trials can't tell us how common that is, because most of them stopped watching after two weeks.²

3How common is it?

It depends on who you ask and how, and the two best-known answers are far apart.

The trial view. A 2024 analysis of 79 studies and about 21,000 people found that 31 percent had at least one symptom after stopping an antidepressant. So did 17 percent of people who stopped a placebo without knowing it. That leaves about 15 percent, or 1 in 6 or 7, whose symptoms came from stopping the drug itself, and about 3 percent with severe symptoms.¹ A 2025 analysis of 50 trials found people who stopped had, on average, one more symptom than people who stayed on or took placebo. Dizziness was the standout, at 7.5 percent versus 1.8 percent.² That placebo figure is the nocebo effect at work, which I wrote about in A fair test.¹³

The long-term view. Most of those trials watched people for two weeks or less after stopping, many had treated them for only a few months, and most stopped the drug abruptly or over a week.²¹ People in real life take antidepressants for years. In a 2025 survey of 310 people in English primary care, those who had used an antidepressant for more than two years were far more likely to report withdrawal, and 64 percent of them called it moderate or severe. Among people who'd used one for under six months, 73 percent reported none or mild.³ That survey had an 18 percent response rate, so people with a bad experience were probably more likely to answer. An earlier review of mostly survey data put the overall figure at 56 percent.¹⁴

I think both are telling the truth about different people. For someone who took sertraline for six months, the trial numbers are a fair guide. For someone who's taken venlafaxine for eight years, they're probably too rosy. Yes, most people do fine, and the people who don't are not imagining it.

4Who gets it, by medicine

Four things raise the odds:

• The medicine. Venlafaxine, desvenlafaxine and imipramine had the highest rates in pooled data, and paroxetine ranked high for severity. Fluoxetine and sertraline had the lowest.¹ In the 2025 analysis, about 20 percent of people stopping venlafaxine had dizziness; vortioxetine produced less than one extra symptom.²
• How long you've taken it. Longer use, more risk.³
• How fast you stop. Abrupt stops and one- to two-week tapers do worst.¹⁵⁶
• Your own history. If missed doses or a past attempt made you feel ill, expect the same again and plan a slower taper.¹⁶

5Withdrawal or relapse?

This is the question that derails more tapers than any other. Someone feels awful a few days after a dose cut, concludes "I guess I really do need this medicine," and goes back on it for another five years. Sometimes that's right. Often it was withdrawal, and a slower taper would have worked.

In the 2025 trial analysis, low mood was not more common after stopping than on placebo in the first weeks, which suggests that depression showing up later is more likely the illness returning than withdrawal.² Anxiety, irritability and tearfulness can be either, which is why timing matters so much.

6Should I stop, and when?

This is a real decision with a real tradeoff, and it's yours to make with good information.

The clearest trial is ANTLER. It took 478 people in UK primary care who'd been on an antidepressant for at least nine months, felt well, and were ready to consider stopping. Half tapered onto placebo over two months without knowing it. Within a year, 56 percent of those who stopped had relapsed, compared with 39 percent of those who stayed on.¹⁹ Read that both ways. Stopping raised the risk. And 44 percent of people who stopped stayed well, as did only 61 percent of those who continued.

How you stop changes those odds. A 2025 analysis of 76 trials found that slow tapering (over more than four weeks) combined with psychological support prevented relapse about as well as staying on the medicine, and prevented about one relapse for every five people compared with stopping abruptly or quickly.⁶ The evidence for the therapy part was rated low, and withdrawal symptoms were poorly reported, so I hold it as encouraging rather than settled.

Things that argue for waiting:

• You've been well for less than six to nine months, or this was your third or later episode. Guidelines generally suggest longer treatment in both cases.⁵⁹
• Life is about to get harder: a move, a new baby, a loss, winter if winters are bad for you.
• Your last episode was dangerous: a suicide attempt, a hospital stay, psychosis.

And things that argue for trying: you've been well for a good while, side effects are costing you something, you're not sure the medicine is still doing anything, or you simply want to know. Those are all legitimate reasons.

7Why the last steps are the hardest

Here's the piece of pharmacology that explains most failed tapers. The effect of an SSRI on its target, the serotonin transporter, doesn't rise in a straight line with dose. It rises steeply at low doses and then flattens. PET scans show that a small dose already occupies most of the transporters, and a standard dose only adds a little more.²⁰

Run that in reverse and you see the problem. Going from a full dose to half barely changes the effect. Going from a quarter dose to nothing takes away most of it at once.¹⁵ So equal cuts in milligrams produce unequal cuts in effect, with the biggest one saved for the end.

This idea, sometimes called hyperbolic tapering, makes good pharmacological sense and matches what patients report. It hasn't yet been tested against other schedules in a large randomized trial, and that's worth knowing.²¹ It's now reflected in UK guidance, which recommends reducing by a proportion of the previous dose and using liquids for the smallest steps.⁹²²

8How to taper

1. Cut by proportion, not by a fixed amount. Reasonable starting points are 25 percent of the current dose every four weeks, or 12.5 percent every two weeks.⁵After years of use, or if earlier attempts went badly, 10 percent a month is a common slower pace.¹⁷
2. Wait before each step. Two to four weeks lets the body catch up and lets any symptoms show themselves.
3. Let symptoms set the pace. Don't take the next step while the last one is still bothering you.⁹
4. If a step is too much, go back one. Return to the last dose that felt fine, steady yourself, then try a smaller cut. Drug labels say the same.⁷²³
5. Go below the smallest tablet. The final steps often need doses smaller than anything sold as a pill.
6. Add support. Therapy during a taper, especially CBT or mindfulness-based approaches focused on staying well, lowers the risk of relapse.⁶⁵

Expect this to take months, not weeks. Labeled tapers are often much faster: the venlafaxine label suggests cutting 75 mg a week, which is too quick for many people, and the same label acknowledges that some need several months.⁷ In a UK trial where family doctors simply reviewed long-term users and supported a gradual taper, more than 40 percent were off their antidepressant at six months.²⁴

9Getting small doses

The practical obstacle is that pills come in a few sizes and a good taper needs many. These are the usual ways around it. Your pharmacist is the best person to ask which fit your medicine.

• Liquids. Fluoxetine, sertraline, citalopram, escitalopram and paroxetine all come as oral solutions or suspensions in the US, which makes any dose possible with an oral syringe.
• Splitting tablets. Fine for plain, immediate-release tablets. Not for extended-release or enteric-coated ones.
• Compounding. A compounding pharmacy can make capsules or liquids in custom strengths. This is often the answer for desvenlafaxine and duloxetine.
• Capsule beads. Extended-release venlafaxine capsules contain small beads or mini-tablets, and some people reduce by removing a counted number. It's off-label, products differ, and it should be done with a pharmacist's guidance.
• Tapering strips. Pre-packed daily doses that step down gradually, developed in the Netherlands. In an uncontrolled study of 895 people, most of whom had failed earlier attempts, 71 percent completed the taper, and two thirds of those followed up were still off one to five years later.²⁵²⁶ They're hard to get in the US.

Skipping days to "stretch" a dose is not a good method for most antidepressants. With a short half-life it produces a daily seesaw of levels, which can cause the very symptoms you're trying to avoid.¹⁵

10If you get stuck

• Reinstate, then slow down. Going back to the last comfortable dose usually settles symptoms within days.⁸ It isn't failure, and it's useful information: the next cut should be smaller.
• Check the basics. Missed or late doses of a short half-life medicine can cause withdrawal symptoms on their own. A once-daily extended-release form, taken at the same time each day, helps.¹⁶
• Hold longer. Some people need six to eight weeks at a step, particularly near the bottom.
• Treat symptoms for a short time. Something for nausea or sleep for a few days is reasonable. Nothing has been shown to work as a specific antidote.
• Consider a fluoxetine bridge. For a stubborn venlafaxine or paroxetine taper, a prescriber can add a low dose of fluoxetine, stop the first medicine after about a week, then taper the fluoxetine, whose long half-life cushions the landing. It's off-label, the evidence is mostly clinical experience, and there are interactions to manage: fluoxetine raises venlafaxine levels, and the two together carry a small risk of serotonin toxicity.⁸¹⁶

11For prescribers

• Say the word "withdrawal" before the taper starts. Patients who know what to expect are less frightened by it and less likely to mistake it for relapse.
• Ask about missed doses. How someone feels after forgetting two days is the best available predictor of how the taper will go.
• Write proportional steps. A share of the current dose every two to four weeks, with explicit permission to hold or step back.⁵⁹
• Prescribe the formulation the taper needs. Liquids and compounded doses for the final steps are part of the plan, not a favor.
• Schedule contact around each step, and check mood, suicidal thoughts and, with SNRIs, blood pressure.⁷
• When symptoms follow a cut within days, reinstate before you re-diagnose. A fast response to the previous dose points to withdrawal.⁸
• Offer therapy alongside. It's the part of the plan with trial evidence for preventing relapse.⁶
• Know the exceptions. Serotonin toxicity, mania or a serious adverse reaction can require stopping fast. MAOIs need specialist care on the way down, and dietary and drug restrictions continue for two weeks after the last dose.
• Don't use a switch to desvenlafaxine as a bridge off venlafaxine. Withdrawal has been reported with that switch.²³

12Where I land

For years my field told patients that stopping an antidepressant meant a week or two of mild symptoms. That was true for many and wrong for some, and the people it was wrong for were often told their symptoms were their illness coming back. We can do better than that, and without swinging to the opposite error of telling people these medicines are traps. They help a lot of people, staying on one is a reasonable choice, and so is coming off.

If you want to try, the plan is not complicated. Pick a calm season. Tell the people close to you. Cut by a share of the dose, wait, and listen to your body before the next step. Make the steps smaller as you go down. If one is too much, go back up one and try again more gently. Most people get to the bottom of the stairs.

How this post was made

I use AI tools in building this site, and I'd rather be open about how. This post began with a question from my clinic and an evidence search I ran in OpenEvidence, a search tool for clinicians that cites the medical literature. I set the scope and the clinical judgment. Claude helped research the wider literature, draft the text and build the tools, and I checked the numbers against the sources below. Any errors are mine.

References