Antidepressants are very common medications, yet their potential negative effects tend not to be adequately discussed with patients. One of those effects is withdrawal, which encompasses an array of uncomfortable and distressing symptoms that can manifest when someone stops taking an antidepressant in any of the common drug classes, including selective serotonin reuptake inhibitors (SSRIs), serotonin-norepinephrine reuptake inhibitors (SNRIs), and tricyclic antidepressants (TCAs). The withdrawal effects can manifest whenever antidepressants are used, regardless of condition, so it can occur during treatment for depression, anxiety, panic disorder, and more.
There are at least a couple factors contributing to the lack of awareness patients (and sometimes even doctors) have regarding antidepressant withdrawal. One is that physicians might be worried that extensive discussion of the negative effects of antidepressants will reduce the likelihood of a patient accepting treatment, while another factor is that physicians are not always acceptably familiar with the risks. As a result, withdrawal from these drugs has been known about for decades, yet it is still under-discussed.
Between 2011 and 2014 in the US, ~13% of people over the age of 12 used an antidepressant in any given month (CDC, 2017) and the average duration of antidepressant therapy has grown to become twice as long as it was in the mid-2000s in the UK (NHS Digital, 2017) and in the US (Mojtabai, 2014). More generally, antidepressant use more than doubled across Western countries between 2003 and 2013 (McCarthy, 2013) such that millions of people are using them on a regular basis and there are millions of people at-risk of experiencing withdrawal.
The rise of antidepressant use has not coincided with a proportionate rise in knowledge about their risks and benefits, contributing to treatment non-adherence. Within three months of starting an antidepressant, up to 45-60% of patients stop taking their medication (Harvey, 2014), a decision that is mostly driven by side effects and ineffectiveness (Samples, 2015). Physicians are more familiar with withdrawal nowadays compared to in the 1990s and 2000s (Young, 1997 ; Haddad, 1999), but there is still room for improvement.
A survey of 1829 patients in New Zealand found only 1% of patients recalled being informed about antidepressant withdrawal when their medication was prescribed (Read, 2018).
Strong, open communication between physicians and patients is very important. Patients who discuss adverse effects with their prescriber are significantly less likely to discontinue therapy (Bull, 2002). It is generally reasonable to suspect that the more patients know about the medication they are given, the less likely they are to feel blindsided and confused if an adverse effect appears. Patients can be so displeased with how their therapy is being handled that they drop out of communication with their physician and cease treatment on their own, a choice known to be associated with a higher risk of withdrawal. Ceasing communication with one’s physician also likely comes with poorer outcomes due to the frequency with which a patient could find a drug or non-drug therapy that works for them if they were to continue working with their prescriber.
Bosman (2016) published a study using information from 38 patients and 26 physicians. Although the physicians felt they provided adequate guidance during withdrawal, many of the patients disagreed and believed the physicians were lacking in the knowledge or time to help. Patients usually reported being the ones to initiate the treatment discontinuation discussion with their physicians, but they wished physicians would take up that responsibility.
Patient noncompliance is common in psychiatry, so it is easy for patients to accidentally experience withdrawal just by missing a dose or two, especially with short-acting drugs like paroxetine or venlafaxine. Demyttenaere (1998) reported only 42.5% of patients had good compliance (80%+ correct intake of medication), while 28.7% had partial compliance (50-80%), and 28.8% had poor compliance (<50%). Noncompliance during psychopharmacological therapy is thought to affect 20-50% of patients (Olivier-Martin, 1986 ; Young, 1986). For 33% of non-compliant patients, discontinuation occurred without their physician being informed (Maddox, 1994). Maddox also found the majority of patients stopped taking their antidepressant within 12 weeks.
A qualitative examination of patients discontinuing or continuing antidepressant therapy was published by Geffen (2011). In that study, people choosing to continue treatment more often had requested SSRI therapy themselves, while the discontinuers had a more negative view of SSRI therapy, more concerns about side effects, and they generally preferred to handle their problems without medication. Continuers reported being more informed about the treatment and its side effects, while discontinuers weren’t satisfied with the information they had been given. Sometimes the prescriber for these patients had to argue with the patients to get them to start therapy and in those cases a simple biochemical explanation (e.g. about a lack of serotonin) was fairly effective for getting people to accept treatment. Among the discontinuers, 4 had a relapse of symptoms and realized they needed to continue SSRI therapy for a longer time.
A survey of 1411 people in the US who reported using antidepressants in the past year found 22% had discontinued treatment without physician advice or approval (Samples, 2015). Patients were at a greater risk of self-discontinuation if they received their prescription from a physician other than a psychiatrist as well as if they had anxiety or substance use disorders. The top reasons for discontinuation were side effects and inefficacy.
Even though detailed patient information leaflets must be provided with all medications in the EU, a study by Haw (2011) looked at 42 leaflets covering 21 antidepressants and found they gave side effect information in a very heterogeneous and not easily understandable way. Some of the leaflets did not even discuss withdrawal and half of them did not provide info regarding how the antidepressant is supposed to work.
In the early days of antidepressant use it was sometimes believed they came with a negligible risk of withdrawal and then once withdrawal was recognized as a common effect, there was a push to use the term “discontinuation syndrome” instead. We know the neuroadaptations that produce antidepressant withdrawal can be grouped alongside the adaptations responsible for withdrawal from drugs of abuse, such as opioids, so it has never made sense to pick a different term (a euphemism, really) for antidepressants.
The SSRI discontinuation syndrome was first defined by an Eli Lilly-funded “Discontinuation Census Panel” that met in 1996 (Schatzberg, 1997). The panel decided antidepressants were associated with a syndrome that is typically mild and self-limiting, but which can impair normal functioning. It was defined as having a duration of 2-3 weeks and it was treated as different from the withdrawal caused by benzodiazepines and other sedatives, even though there was no evidence to support the distinction.
As Davies (2018) notes, the use of “discontinuation” in place of “withdrawal” is not just a euphemism, it is misleading because the symptoms can appear without actual treatment discontinuation, such as between doses or after one or two missed doses.
Because withdrawal is the standard term applied to other drugs, it only makes sense to use it when discussing antidepressants. It is important to understand that producing withdrawal is not the same as producing addiction. While people do adapt to these drugs in a way that produces physical dependence, dependence is not the same as addiction, which involves the continued use of a drug despite negative consequences, typically in an “abusive” manner. People do not abuse antidepressants and they do not pursue their medication in an addictive way. They are often given to people who have a history of addiction, yet there are essentially no reports of people abusing typical antidepressants. Antidepressants are also not self-administered by animals, a sign that they lack reinforcing qualities (Haddad, 2005).
Nevertheless, these drugs have been popularly viewed as “addictive” for decades. Priest (1996) reported 78% of the public believed antidepressants were addictive and only half of nurses had a firm belief that they were not addictive (Gray, 1999).
It is quite common for drugs to cause physical dependence and withdrawal while not producing addiction. This occurs with anticonvulsants, β-blockers, diuretics, antihypertensives, antipsychotics, and more (Haddad, 2005).
The only exception to this is with some MAOIs, which are occasionally taken in a more abusive way. For example, Le Gassicke (1965) reported a case where a 34-year-old male had psychological and physical dependence on tranylcypromine. He had been using 200 mg/d on average and would sometimes take up to 700 mg/d. When he finally got off the drug he exhibited withdrawal symptoms such as fatigue, weakness, drowsiness, depression, visual distortions, vivid daytime dreams, and nightmares.
Part of the reason physicians and patients are inadequately informed about antidepressants is that the scientific literature is biased. Even if a prescriber wants to share information with a patient, they are reliant on a collection of studies that often paint a misleading picture to downplay or ignore the risks. Like with other kinds of medications, we have evidence showing there is a “file drawer” effect with antidepressant studies, meaning positive results are more likely to be published so that it appears as if the drugs are more effective and less prone to side effects than they really are.
Data from the FDA shows second-gen antidepressants (e.g. SSRIs) are effective based on both the published and unpublished literature, but the distribution of results is heavily skewed (Vries, 2016). 94-96% of published trials are positive compared to only 51% of all (published + unpublished) depression trials and 72% of anxiety trials. Inadequate reporting of negative effects has been an issue with SSRIs and TCAs, and it is likely a factor with all drug classes.
Vries (2016) found FDA reviews and journal articles were lacking in discussion of serious adverse effects (SAEs). Out of 97 articles covering 102 trials, just 37% mentioned SAEs. And of the 21 articles that provided study-level SAE information, only 6 included complete reporting of SAEs. There was a tendency for the published articles to describe SAEs in a way that downplayed their significance, such as a case where an adverse effect was listed as “emotional lability” even though the case involved “emotional lability/intentional overdose of paracetamol and aspirin.” In cases where the articles were missing descriptions for the SAEs, the FDA found the negative effects were primarily psychological, such as suicide attempt, depression, and mania.
This data suggests some of the effects patients are worried about and that they may encounter during therapy are not properly discussed in the literature and it is unlikely physicians are fully informed. This can make people less likely to stick with therapy and it can be a trigger for abrupt discontinuation, which comes with the greatest risk of withdrawal.
Also, beyond classic biases, another issue is simply that antidepressant studies have not historically had a withdrawal phase, leaving us with very little data to go on.
Effects and Occurrence
Withdrawal is not immediately a risk when someone takes antidepressants, but by the time ~6 weeks of use has occurred, it is possible a patient will experience negative effects upon stopping. The response to withdrawal is highly variable, with some patients experiencing little to no discomfort even with abrupt discontinuation while others experience severe, debilitating symptoms for weeks or months (Price, 1996). Research shows a typical scenario is a 1-6 week period of mild/moderate withdrawal effects.
The risk can be reduced by a prolonged taper in the realm of months, not weeks (Taylor, 2012), though there is a small minority who will still experience negative effects during dose reduction or treatment cessation regardless of the taper duration. Why the experience differs so drastically between patients is unknown and that is why people need to be prepared for the worst scenario even if they are unlikely to experience is.
Withdrawal from SSRIs/SNRIs can cause electric shock sensations (e.g. “brain zaps”), flu-like effects, nausea, tiredness, dizziness, insomnia, anxiety, vivid dreams, and irritability (Keks, 2016). Stopping TCAs is associated with similar effects, but with a greater incidence of GI problems like diarrhea and pain. And MAOI withdrawal is typically the most troublesome, coming with effects like agitation, irritability, mood issues, cognitive impairment, and sometimes psychosis or delirium.
During the withdrawal period there can be a return of a patient’s original symptoms as well as an exacerbation of those symptoms. When depression returns shortly after stopping therapy it should be considered an effect of the withdrawal, not a relapse of the original condition, since depression relapse usually takes longer to present. It is important not to jump to the belief that the return of depression is a relapse since this could result in an inappropriate return to therapy.
Because the effects are usually mild and self-limiting, treatment is not required, but there are scenarios where the symptoms are severe enough to warrant a return to the original antidepressant, the initiation of an easier to manage antidepressant, or the use of a longer taper.
A “persistent post-withdrawal disorder” characterized by an unusually long period of withdrawal effects for months and occasionally years has been recognized (Belaise, 2014), but it is far from the norm.
The new and increased symptoms reported by 28 SSRI or venlafaxine patients studied by Tint (2007) included: dizziness (42%), headache (42%), nervousness/anxiety (42%), panic/sudden anxiety (32%), agitation (32%), nausea (32%), and sudden worsening of mood (32%). Depressive symptoms also increased significantly and 4 of 8 paroxetine patients had emergent suicidal ideation. Symptoms, including the depressive effects, basally returned to baseline either upon starting an antidepressant again or after a week of unassisted withdrawal.
Not all withdrawal effects are truly caused by a medication, as shown by studies in which stopping placebo administration also causes withdrawal-type effects, though at a much lower rate (~10%) (Oehrberg, 1995).
Disequilibrium and sensory abnormalities appear to be more common when withdrawing from SSRIs vs. TCAs, but TCAs cause more GI symptoms (Haddad, 2005).
In a study of 1051 patients given escitalopram, 336 given paroxetine, 124 given venlafaxine XR, and 239 given placebo, all antidepressants produced significantly more withdrawal effects (Baldwin, 2007). Fewer symptoms occurred with escitalopram compared to paroxetine or venlafaxine XR given for depression, and it also produced fewer symptoms in cases of social anxiety and general anxiety disorder (GAD). Overall the effects in depressed patients were similar to those in anxiety patients. The effects peaked after a week and were substantially lower within 2 weeks, but some effects occasionally persisted beyond that point.
Occurrence, Onset, and Duration
Most people do experience some kind of withdrawal upon stopping treatment. Double-blind, placebo-controlled studies show around half of patients abruptly stopping antidepressants after using them for at least several months will have one or more symptoms (Fava, 1997 ; Rosenbaum, 1998). With SSRIs the rate of withdrawal ranges from 42-100% with paroxetine and from 9-77% with fluoxetine (Rosenbaum, 1998 ; Bogetto, 2002). Based on 14 studies, the mean rate overall was 54%, with differences between the drugs potentially coming from different half-lives, receptor affinities, and tapering methods (Horowitz, 2019).
An online survey of 817 patients conducted by the Royal College of Psychiatrists found 63% reported some degree of withdrawal (Davies, 2018). The risk may be higher when people have taken their medication for at least a few years, as was seen in a survey of 1367 patients in New Zealand. Of those patients, 55% reported withdrawal, but that increased to 74% when they had taken an antidepressant for over three years (Cartwright, 2016). However, the literature does not reliably show a large effect of multi-year treatment duration compared to shorter durations and it appears likely a cap on the risk and severity will usually be reached somewhere between six months and a few years, but more research is needed on this point.
Similar rates have been reported with other antidepressant classes, such as 21% to 100% with the TCA imipramine (Lejoyeux, 1996).
On average, the mean time to symptom onset is 1-3 days (Bogetto, 2002).
The onset differs between drugs. Those with a short half-life, such as paroxetine and venlafaxine, are associated with faster symptom onset, greater severity of effects, and possibly a higher risk of withdrawal in general. When a long-acting drug like fluoxetine is used, stopping the medication causes a slower disappearance of the drug from the user’s body such that a form of tapering takes place for 1-2 weeks, allowing patients to be exposed to progressively less medication. Among 97 patients discontinuing paroxetine or fluoxetine after a successful 8-week course of treatment, 27% reported withdrawal, with 85% of those individuals having been prescribed paroxetine compared to 15% who were prescribed fluoxetine.
Half-life differences are important, but they are not the whole story. For example, Price (1996) found withdrawal was significantly more common with paroxetine compared to fluvoxamine even though those SSRIs have similar half-lives.
A review of 22,422 calls to the Maudsley medication helpline in the UK between 1997 and 2005 found 7.8% (n=1753) were about antidepressant withdrawal (Taylor, 2006). Paroxetine was involved in 40% of cases and venlafaxine was involved in 14%, while no other individual drug contributed over 10%.
The 1-2 week duration often cited in the literature is not well-supported. Based on a study by the Royal College of Psychiatrists, symptoms usually lasted up to 6 weeks and a quarter of patients reported symptoms for more than 12 weeks. Another report found 87% of patients experienced at least two months of symptoms and 59% had a year of effects, while 16% reported a duration greater than three years (Davies, 2018). These figures are outliers based on the literature as a whole, but they demonstrate prolonged withdrawal should be treated as a relevant risk.
Patients occasionally report problems upon switching to a generic version of their medication even though generics are regulated to be effectively equivalent to branded drugs. There is little research on how this affects the risk of withdrawal specifically, but taking a wider view of the subject of branded vs. generic antidepressants suggests they are essentially always the same. For example, an analysis of 16,000 new SSRI/SNRI users found initiating treatment with a branded drug vs. a generic was not associated with superior outcomes (Vlahiotis, 2011). And an issue of the Carlat Psychiatry Report in 2009 found most claims of alleged non-equivalence were from “single cases or very small case series, virtually all written by authors who are also paid consultants for pharmaceutical companies” (Carlat, 2009).
Patients should not expect to experience a change in their medication’s kinetics or efficacy that produces withdrawal when they switch from a branded to generic drug.
More Severe Psychological Changes
An acute exacerbation of mood issues is a risk during the withdrawal period, which may be responsible for the 60% increase in suicide attempts that is observed in the 14 days following cessation (Horowitz, 2019).
A lot has been published on manic and hypomanic states triggered by the withdrawal from all common antidepressant types (Narayan, 2010), though it is important to ensure the right diagnosis is made in these situations since withdrawal-induced mania could actually be mania caused by antidepressant use, mania in an existing bipolar patient, or mania from other drug use and/or other drug withdrawal. It is also possible for patients to experience agitated depression with symptoms like irritability or agitation, which can look somewhat like hypomania (Narayan, 2010).
The first published case series appears to have been from Mirin (1981), who described 1 bipolar patient and 6 unipolar patients who had hypomania (n=5) or mania (n=2) within days of stopping a TCA, mostly amitriptyline. Although three abruptly stopped taking their medication, the other four tapered over a 2-8 week period.
In a literature review by Narayan (2010), 42 mania/hypomania reports were found and 24 clearly met the study’s inclusion criteria. Most cases included unusually high mood, but otherwise it was not possible to determine whether the cases represented mania or hypomania based on the published information. It was considered likely most of the cases were hypomanic episodes. In 9 cases no treatment was given and the symptoms resolved in a median of 26 days. In the 6 cases where an antimanic treatment was given, all patients improved.
Mania/hypomania typically resolve upon restarting the antidepressant and it does not appear patients experiencing this effect have a genetic predisposition as per family history (Ali, 2003).
Serotonergic and noradrenergic changes have been hypothesized to play a role in mania or hypomania during discontinuation, but so far there is little research to support any mechanism.
Because antidepressants cross the placenta, they can produce withdrawal in neonates. An early report of the syndrome was from Chambers in 1996 and it has since become a commonly discussed side effect of antidepressant use. Symptom onset is between birth and a few hours after birth, and usually the symptoms are just mild and self-limiting, with a duration of <2 weeks (Haddad, 2007).
Symptoms: shivering, tremor, restlessness, increased muscle tone, respiratory difficulty (e.g. tachypnea or cyanosis), irritability, crying, disrupted sleep, and feeding trouble. Rarely the symptoms are severe, such as seizures or hyperpyrexia (Haddad, 2005).
Because 10-16% of pregnant women meet the criteria for depression (Andersson, 2004 ; Bennett, 2004 ; Newport, 2001), it is important for antidepressants to be available, but the effect on neonates should be kept in mind as well. Antidepressants are used by 0.1 to 1.8% of pregnant women during the first trimester (Andrade, 2004 ; Rubin, 1993 ; Buitendijik, 1991 ; Egen-Lappe, 2004). Cohen (2006) reported half of pregnant women who stop their antidepressant prior to conception end up having to restart it due to depression during pregnancy. Antidepressant therapy during pregnancy appears to be effective, with medication use being associated with a 30% lower relapse rate per month vs. patients who have discontinued (Viguera, 1998).
Kallen (2004) found TCA and SSRI use during pregnancy produced a significantly higher rate of neonatal respiratory distress, hypoglycemia, and convulsions, with no difference between drug classes. Though Kallen notes many of the patients were on additional drugs as well, so antidepressants may not be the only factor.
Diamond (1989) and Garner (1993) estimated 20-50% of neonates could develop TCA withdrawal due to maternal use.
A study of 16,000 pregnancies found 14% of SSRI-exposed neonates had respiratory distress compared to 8% of unexposed neonates, and the exposed neonates also had more feeding problems and jaundice (Oberlander, 2006). Compared to unexposed neonates, Laine (2003) reported neonates exposed to SSRIs had a 4-fold higher risk of serotonergic symptoms like tremor, restlessness, and rigidity, although these effects might be related to a lasting pro-serotonergic effect of the SSRIs, not due to withdrawal.
Neontal exposure to antidepressants does come with some risks, but usually the withdrawal effects are mild and transient. Exposure to these drugs poses relatively little trouble and it does not seem to cause any long-term negative effects (Nulman, 2002 ; Nulman, 1997 ; Gentile, 2005).
Symptoms of SSRI withdrawal include dizziness, nausea, vomiting, headache, lethargy, ataxia, vertigo, electric shock sensations, paresthesia, aggression, impulsiveness, and suicidality (Tamam, 2002). In most cases the symptoms appear within 1-3 days of cessation or dose reduction and the chance of withdrawal is high (around half of patients, like with other classes), though it varies between drugs. Paroxetine appears most likely to trigger withdrawal symptoms (Tamam, 2002) but symptoms have been reported after stopping every SSRI.
Among 74 patients who recently discontinued an SSRI, 21% did so abruptly, and there was a significantly greater number of discontinuation symptoms (12 vs. 5.0) in patients who abruptly stopped compared to those who gradually stopped (Van Geffen, 2005).
Rosenbaum (1998) conducted a study of 242 patients with remitted depression who had received fluoxetine, sertraline, or paroxetine for 4-24 months. They had their medication use interrupted with placebo substitution for 5-8 days. The mean increase in withdrawal symptoms was significant for sertraline and paroxetine patients, but not for those given fluoxetine. And the number of symptoms was also significantly lower with sertraline patients vs. paroxetine patients. Fluoxetine patients did not have a significant increase in anxiety, depression, or anger/hostility during the substitution period, unlike those given paroxetine or sertraline. Rosenbaum found the only symptom spontaneously reported by over 10% of fluoxetine patients was headache, while four symptoms (dizziness, headache, nervousness, and nausea) were reported by over 10% of sertraline patients and eight symptoms (dizziness, nausea, insomnia, headache, abnormal dreams, nervousness, asthenia, and diarrhea) were reported by over 10% of paroxetine patients.
Black (2000) reported symptom onset was 1-3 days for 81% of patients with withdrawal and 94% had an onset within a week. Based on an older retrospective review of 171 patients, symptoms persisted a mean of 12 days and a max of 3 weeks, regardless of which SSRI was used (Coupland, 1996). However, we now know withdrawal can last much longer.
Usually withdrawal only occurs if someone has been consistently using an SSRI for at least four weeks. Coupland (1996) reported no incidence among people who were treated for under seven weeks, meanwhile on the other end, increasing treatment duration beyond six months did not significantly increase withdrawal risk.
Withdrawal is possible with all SSRIs, but the incidence is usually highest with paroxetine, while it is lowest with fluoxetine and escitalopram. Fluoxetine’s long half-life, coupled with the even longer half life of its active metabolite norfluoxetine, contributes to the reduced withdrawal burden.
In a placebo substitution study, withdrawal become evident in paroxetine patients as early as the second dose of placebo (Michelson, 2000). Among 87 depressed patients given paroxetine, fluoxetine, sertraline, or citalopram, withdrawal was reported upon discontinuing each drug, but it was worst with paroxetine (Belaise, 2012). In a 12-week trial of paroxetine and escitalopram for anxiety, discontinuation produced no significant effects in the escitalopram patients, while paroxetine patients had significantly greater symptoms (Baldwin, 2006).
Among 97 outpatients treated for dysthmia for 8 weeks or longer with paroxetine or fluoxetine, withdrawal was seen in 26 (27% of the sample), 22 of whom had been on paroxetine and 4 of whom had been on fluoxetine (Bogetto, 2002).
A retrospective review of 385 people treated with paroxetine found withdrawal symptoms in 11% of patients and in essentially every case restarting paroxetine resolved the symptoms (Himei, 2006). In those patients, a slower paroxetine taper was then used, allowing for successful discontinuation without a return of symptoms. Himei also reported the symptoms were much more common in patients who abruptly stopped vs. those who tapered under supervision.
The Discontinuation Consensus Panel funded by Eli Lilly found SSRI withdrawal caused somatic and psychic symptoms:
- The somatic effects were in five clusters: disequilibrium (e.g. dizziness, vertigo, and ataxia) ; GI problems (e.g. nausea and vomiting) ; flu-like effects (e.g. fatigue, lethargy, muscle pain, and chills) ; sensory disruption (e.g. paresthesia, electric shocks) ; and sleep trouble (e.g. insomnia and vivid dreams).
- Psychic symptoms: anxiety, agitation, crying, and irritability.
In a study of 46 patients (30 given paroxetine, 8 given sertraline, 5 given fluoxetine, and 3 given fluvoxamine) 58% had a spontaneous resolution of withdrawal symptoms, while 38% required an increase in dose or medication reintroduction (Black, 2000). 48% reported symptom resolution within a week, but other patients required up to 13 weeks for resolution. Whenever the SSRI was restarted or increased, symptoms resolved within 72 h in all cases and within just 24 h in most cases.
Hindmarch (2000) studied the cognitive effects of abruptly switching patients from an SSRI to placebo for 4-7 days. Paroxetine patients generally had a larger number of cognitive failures compared to fluoxetine and citalopram patients. Paroxetine withdrawal was associated with significantly worse incoordination relative to other drugs. Sleep quality was also affected, with paroxetine patients reporting impaired sleep quality that was worse than those withdrawing from sertraline or citalopram. The increase in depression during the switch was also greatest in people previously on paroxetine.
The difference between fluoxetine/escitalopram and paroxetine is pretty consistent, but its magnitude varies between studies. In some, fluoxetine (14% rate) barely causes withdrawal relative to paroxetine (66% rate) or sertraline (60% rate) (Rosenbaum, 1998). And when withdrawal does occur after using fluoxetine, the effects are significantly reduced compared to the other antidepressants. Rosenbaum (1998) also reported no change in depressive symptoms during withdrawal from fluoxetine, whereas sertraline and paroxetine patients had a significant increase.
An analysis of spontaneous adverse drug reaction reports by UK doctors showed discontinuation rates per 1000 prescriptions of 0.3 for paroxetine, 0.03 for sertraline, 0.003 for fluvoxamine, and 0.002 for fluoxetine (Price, 1996). And even though fluoxetine was the most common antidepressant in Australia during the mid-1990s, the largest number of discontinuation reaction reports were for paroxetine (ADRAC, 1996).
A large analysis of 49,393 case reports collected from 47 countries by the World Health Organization (WHO) found there were 947 withdrawal reports from paroxetine, 271 from fluoxetine, and 170 from sertraline (Stahl, 1997). Most cases gave limited information about the effects, but the median time to onset was 2-3 days after stopping.
Coupland (1996) reviewed case notes on patients who stopped their antidepressant under supervision and found the incidence of withdrawal was much higher with clomipramine (31%) or with a short-acting SSRI like fluvoxamine or paroxetine (17%) than with a longer-acting drug like sertraline or fluoxetine (2%).
In an open-label retrospective study of 171 patients undergoing SSRI taper or withdrawal over a 2-week period or 4-week period, withdrawal occurred in 31% of clomipramine patients, 20% of paroxetine patients, 14% of fluvoxamine patients, 2.4% of sertraline patients, and 0% (n=0/20) of fluoxetine patients (Coupland, 1996).
Among 120 panic disorder patients given either placebo or paroxetine for 12 weeks, an abrupt change to placebo for 2 weeks produced withdrawal in 35% of people given paroxetine and 14% of those given placebo (Oehrberg, 1995). Judge (2002) compared the effect of placebo-substitution in 75 fluoxetine patients vs. 75 paroxetine patients. After a mean of 4 days from discontinuation, the groups were significantly different, with significantly elevated withdrawal effects in the paroxetine patients but not fluoxetine patients. Paroxetine withdrawal also had a significant effect on measures of depression and social functioning, while fluoxetine did not.
Escitalopram may have a preferable withdrawal profile (Baldwin, 2007) that is partly attributable to a slow dissociation rate from SERT (820 min) in the presence of serotonin in vitro. Though in a study of 25 depressed patients given escitalopram for six months before withdrawal, symptoms occurred in 14 patients and they included dizziness (44%), muscle tension (44%), chills (44%), confusion or trouble concentrating (40%), amnesia (28%), and crying (28%) (Yasui-Furukori, 2016). Escitalopram was readministered due to severe withdrawal in two patients. Oddly, treatment dose and plasma concentration were significantly higher (not lower) in patients with withdrawal, suggesting the hypothesis that a rapid decline in antidepressant concentration causes withdrawal might not be entirely accurate.
Rarely hallucinations can occur during withdrawal. Two cases associated with paroxetine discontinuation were reported by Yasui-Furukori (2011). Both patients experienced visual and auditory hallucinations along with more typical withdrawal effects like dizziness, insomnia, nausea, and headache. Restarting paroxetine alleviated the symptoms. And in a separate report a middle-aged female experienced delirium, emotional lability, confusion, and hallucinations upon stopping fluoxetine (Blum, 2008).
A publication by Szabadi (1992) described a postpartum patient with obsessive-compulsive disorder (OCD) who had hypomanic symptoms and aggression 1 to 2 days after stopping fluvoxamine treatment. The same patient reported impulsivity and aggression whenever they tried to stop fluvoxamine during the pregnancy.
Besides half-life differences between the SSRIs, another proposed factor contributing to why paroxetine withdrawal is typically worse than average is that paroxetine has a more significant affinity for muscarinic acetylcholine receptors than other SSRIs, so chronic use could alter cholinergic activity (not just serotonergic), similar to some TCAs (Stanford, 1996).
A double-blind study using patients with panic disorder demonstrated adverse weffects (typically mild or moderate) occurred in 35% of patients during a 2-week withdrawal period from paroxetine, which had been given for 12-weeks (Oehrberg, 1995). This was higher than the 14% rate of symptoms in the control group.
The incidence of withdrawal from MAOIs is similar to other antidepressants. Tyrer (1984) reported 32% of people withdrawing from MAOIs experienced withdrawal.
Compared to other antidepressant classes, withdrawal from MAOIs can be more severe, with tranylcypromine and phenelzine appearing to be the worst. Symptoms include depression beyond the original condition, suicidality, acute confusional states with disorientation, paranoid delusions, and hallucinations, along with anxiety, depersonalization, and hyperacusis (Haddad, 2007). A large number of other effects have also been reported over the years, including: pressured speech, aggression, slowed speech, myoclonic jerks, catatonia, ataxia, athetosis, and irritability (Dilsaver, 1994).
The cognitive impairment can present with disorientation towards time and place, disorganized thinking, and difficulty recognizing familiar faces (Dilsaver, 1994).
Psychosis can occur in people withdrawing from phenelzine and tranylcypromine even with no prior history of psychotic illness (Frankel, 1985, Liskin, 1985).
While on tranylcypromine, REM sleep is strongly suppressed and during discontinuation patients report insomnia, fear of being alone, and fear of being in the dark. They also have repeat awakenings through the night and frequently go straight from wakefulness or drowsiness into REM sleep. In 12 of 19 patients there was a direct transition to REM sleep from wakefulness captured by nocturnal polysomnography (Dilsaver, 1994). Patients report vivid dreams due to the sudden rise in REM sleep.
Tiller (1997) reported no discontinuation symptoms in 624 patients who ceased moclobemide (300 to 600+ mg/d) after up to three years of treatment. A review of the safety database of Roche, the maker of moclobemide, also found no mention of withdrawal effects (Hilton, 1995). This may indicate moclobemide has a relatively low propensity to cause withdrawal, although there are still case reports of withdrawal symptoms like cramps, headache, nausea, and hot flushes (Curtin, 2002).
Because MAOI withdrawal can be particularly severe, it needs to completed slowly with a long taper and patients should be carefully monitored.
A literature review investigating SNRI withdrawal found all SNRIs are associated with withdrawal effects, though venlafaxine may be the worst (Fava, 2018 ; Montgomery, 2009 ; Stone, 2007 ; Bhat, 2017). Symptoms usually start within a few days of cessation and can persist for a few weeks and occasionally longer. The effects include anxiety, depression, dizziness, insomnia, nausea, electric shock sensations, emotional lability, and suicidal ideation.
Among 20 patients receiving venlafaxine XR for depression, 78% of venlafaxine patients reported withdrawal during the 3 days after discontinuation, compared to 22% of placebo patients (Fava, 1997). The top symptoms were dizziness or lightheadedness, excessive sweating, irritability, dysphoria, and insomnia.
In an 8-week study of 163 people given sertraline or venlafaxine XR, venlafaxine withdrawal was associated with more moderate to severe symptoms, but the overall qualitative aspect of SSRI vs. SNRI withdrawal-induced symptoms is similar (Sir, 2005).
The withdrawal effects of TCAs can be grouped into five main categories, according to Tamam (2002):
- GI and general somatic symptoms,including anxiety, agitation, muscle tension, nervousness, flu-like symptoms (fatigue, headache, sweating, myalgia), lethargy, nausea, vomiting, asthenia
- Sleep trouble, e.g. insomnia along with excessive, vivid dreams.
- Movement disorders, e.g. akathisia, parkinsonism, unsteady gait, abnormal movements of mouth and tongue
- Behavioral activation, e.g. panic attacks, delirium, mania, or hypomania.
- Misc, e.g. cardiac arrhythmias
Wolfe (1997) reported the incidence of withdrawal with TCAs is 21-80%, the onset is within days, and the duration is days to weeks. The effects are characterized by excess adrenergic activity, mania, flu-like symptoms, sleep trouble, and extrapyramidal symptoms. Among 45 patients stopping imipramine, 85% of those who had been treated for 2 months or longer had withdrawal symptoms, compared to 16% of those treated for under 2 months (Kramer, 1961). Symptom onset was within two days and a large minority (40%) of patients had giddiness as one of their symptoms.
A prospective assessment of 9 patients subjected to abrupt withdrawal from TCAs found 7/9 had withdrawal symptoms (Ceccherini-Nelli, 1993). Four had some combination of reduced appetite, nausea, vomiting, chills, diaphoresis, or malaise. Two patients (one withdrawing from amitriptyline and the other from imipramine) met criteria for hypomania even though neither had a history of hypomania.
All 22 children undergoing withdrawal from imipramine over a mean of 6.4 days reported at least one withdrawal symptom within 10 days of stopping (Law, 1981). Bialos (1982) found 80% of adults prescribed amitriptyline developed symptoms within two weeks of withdrawal.
Sleep disruptions are common, leading to initial and middle insomnia, along with unusually prevalent dreaming. Dilsaver (1994) reported patients often feel as though they are immediately entering a dream once they fall asleep and the dreams are typically vivid and scary.
Anxiety, agitation, jitteriness, nervousness, and even panic attacks have been reported, owing to the “activation” aspect of TCA withdrawal (Dilsaver, 1994).
In a small minority of cases withdrawal produces cardiovascular problems, something not associated with other antidepressant classes. A man who was on imipramine experienced a severe cardiac arrhythmia following withdrawal (Boisvert, 1981). Initially he had nausea, vomiting, and excessive dreaming, and then he complained of palpitations and “skipped beats.” An ECG showed sinus tachycardia, ventricular contractions, bigeminy, and frequent ectopic beats.
A middle-aged female with no history of cardiovascular pathology experienced arrhythmia after abrupt clomipramine withdrawal (Van Sweden, 1988). She was anxious and had insomnia. ECG revealed tachycardia, premature ventricular contractions, and bigeminy.
Some cases of significant mood change following TCA withdrawal in people also receiving lithium have been reported. Ghadirian (1986) wrote of two cases where a TCA was given alongside lithium and within 24 h of stopping the TCA (due to inefficacy) they had significantly elevated mood to the point of mild euphoria, followed by a return to euthymia over a 1-week period.
Three articles describing 11 patients who experienced behavioral activation, hypomania, and/or mania after TCA withdrawal were published in the 1980s (Dilsaver, 1983 ; Mirin, 1981 ; Nelson, 1983). That degree of activation is atypical, but it has been reported in enough cases for it to clearly be a potential effect of TCA withdrawal.
Comparatively little has been published about withdrawal from mirtazapine but it does seem to cause withdrawal symptoms like anxiety, stimulation, elevated mood, insomnia, dizziness, panic, irritability, and more (Cosci, 2016 ; Pombo, 2017). Overall the effects are centered around psychological disturbances and increased stress. Rarely it may trigger hypomania (MacCall, 1999). Like with other classes, restarting the drug quickly resolves the withdrawal effects.
The neurobiological mechanism of antidepressant withdrawal is not fully understood, but we have a fair idea of some of the contributing factors. Common antidepressants typically function by increasing the activity of monoamines, namely serotonin, dopamine, and norepinephrine. Over time this produces neuroadaptations responsible for their efficacy, but another consequence is physical dependence, where hypoactivity or hyperactivity of various systems produces uncomfortable symptoms if someone goes without their medication. These adaptations come to the forefront upon cessation and the faster a drug leaves, the faster the adaptations and their associated effects become evident.
Most antidepressants have an effect on serotonin, either by binding to the serotonin transporter (SERT) to block the reuptake of serotonin or by inhibiting enzymes that break down serotonin, as is the case with MAOIs. This changes the concentration of serotonin and affects the sensitivity and expression of serotonin receptors. Because withdrawal often comes with a general “activating” effect that increases anxiety, insomnia, and stress, serotonin’s control over glutamate (the major excitatory neurotransmitter in the brain) is of interest.
Studies have shown rapid tryptophan depletion in depressed patients treated with fluoxetine leads to greater relapse vulnerability vs. patients treated with the TCA desipramine (Heninger, 1996). Levels of the serotonin metabolite 5-HIAA are increased during withdrawal from chronic fluoxetine in rats and a 30-50% increase vs. control persists for at least 14 days after cessation. During the first 24 h after stopping fluoxetine following a 3-week administration period in animals, serotonin turnover is reduced but then turnover increases significantly after an 8-day washout period.
Activity at some serotonin receptors reduces glutamate release (e.g. 5-HT1A, 5-HT1B, and 5-HT6), though activity at others can increase it, like with 5-HT3. Because serotonin affects glutamatergic activity, a rise or fall in glutamate is to be expected upon stopping a serotonergic antidepressant. Indeed, some researchers have proposed unusually high glutamatergic activity contributes to withdrawal (Harvey, 2002 ; Harvey, 2003) and studies have supported a role of glutamate in stress and depression.
Whether the acute withdrawal period is characterized by a general lack or excess of serotonin activity is unclear. The simplistic, outdated explanation for how serotonergic antidepressants function is that they increase serotonin, so naturally it would be assumed stopping the drug leaves someone with a temporary lack of serotonin. In reality, the literature is much more complicated on this issue to the extent that an argument can even be made for antidepressants functioning via a general decrease in serotonergic activity, not an increase.
5-HT1A autoreceptors, which reduce serotonin release, are downregulated with chronic antidepressant use and changes at 5-HT1A appear important in the efficacy of these medications. Following abrupt withdrawal of a serotonergic medication, it is possible serotonin release will be increased due to a lack of inhibition by presynaptic serotonin near those receptors, triggering a hyperserotonergic state. Alternatively, there may only be a drop in serotonin that is not subsequently linked to increased serotonin release, which is a better explanation for why buspirone, a 5-HT1A agonist that reduces serotonin release, worsens antidepressant withdrawal (Carrazana, 2001).
The serotonergic changes from antidepressants could also influence activity in the acetylcholine, GABA, norepinephrine, and dopamine systems (Zajecka, 1997).
The most common explanation for how serotonin is involved is that withdrawal leaves people in a relatively hyposerotonergic state and that is why a lot of the symptoms are related to processes known to involve serotonin, such as vertigo, gain instability, visual changes, and dizziness. A reduction in activation of 5-HT1A receptors in the raphe and vestibular nuclei is thought to be a likey explanation for dizziness, vertigo, nausea, and lethargy stemming from withdrawal.
In some cases where the 5-HT2 antagonist cyproheptadine is prescribed for SSRI-induced sexual dysfunction, the drug reduces or blocks the SSRI’s efficacy (Therrien, 1997). Yet among these patients cyproheptadine did not trigger any of the symptoms classically associated with withdrawal, so 5-HT2 receptor activity might not play a major role in withdrawal. In the minority of cases where movement disorders like dystonia are triggered, the symptoms are most likely related to dopaminergic activity, according to Horowitz (2019).
For SSRIs, effects at non-SERT targets are likely important, such as paroxetine’s high affinity for muscarinic receptors and appreciable affinity for NET (Owens, 2001 and Tatsumi, 1997), sertraline’s DAT inhibition (Harvey, 1997), fluoxetine’s 5-HT2C antagonism (Sanchez, 1999), and citalopram’s histamine H1 receptor affinity (Harvey, 2000).
After stopping chronic fluoxetine, rats show a significant increase in locomotion, suggestive of withdrawal effects comparable to the agitation and anxiety experienced by some humans (Bjork, 1998). Also, rats chronically given citalopram showed an exaggerated startle response during withdrawal (Bosker, 2010) and serotonin turnover was found to be increased.
In animals, downregulation of SERT by the SSRI sertraline appears to normalize 1 week after stopping, while 5-HT1A autoreceptors are desensitizd for at least 48 h after withdrawal, and it takes three days for the mRNA levels of 5-HT1B to return to baseline. This moderately extended period during which SERT and serotonin receptors return to normal after chronically being exposed to antidepressants demonstrates how it could conceivably take weeks or longer for withdrawal effects caused by serotonergic adaptations to subside.
Harvey (2014) demonstrates plausible contributors to withdrawal based on what the literature says about agomelatine, an atypical antidepressant that has not yet been associated with withdrawal (Chanrion, 2008). While it does increase serotonin in the long-term, it does not acutely increase synaptic serotonin and it lacks appreciable activity at adenosine, adrenergic, dopaminergic, GABA, muscarinic, nicotinic, excitatory amino acid, benzodizepine, and sigma receptors, along with having negligible effects on sodium, potassium, and calcium channels (Guardiola-Lemaitre, 2014). Chronically it does not alter 5-HT1A density and it does not increase cell surface density or sensitivity of 5-HT2C receptors (de Bodinat, 2010 ; Guardiola-Lemaitre, 2014).
Agomelatine’s short 2.5 half-life would be expected to come with a significantly higher risk of withdrawal and of temporary adverse effects following missed doses, yet it still is not associated with withdrawal. The drug is primarily an agonist at the melatonin receptors MT1 and MT2, while also functioning as a 5-HT2C and 5-HT2B antagonist to a much smaller extent. Its antagonism of 5-HT2C receptors on monoaminergic neuron bodies could hypothetically disinhibit dopamine and norepinephrine release in the frontal cortex, contributing to antidepressant effects. MT2 receptors can inhibit nitric oxide activity due to being negatively coupled to guanylyl cyclase (Dhir, 2011), likely contributing to the antidepressant effects and reduced withdrawal potential.
56 Japanese patients with depression or anxiety who received paroxetine for at least 8 weeks were abruptly or gradually stopped, with a mean taper length of 3.4 days (Murata, 2010). Overall, 36% had withdrawal effects, but no significant effect of genetics was identified in terms of polymorphisms affecting 5-HT1A (i.e. -1019C allele and the Gly272Asp polymorphism), 5-HT2A, 5-HT2C , 5-HT3A, 5-HT3B, or SERT and there was also no significant effect of CYP2D6 enzyme genotype. Because the sample size was quite small, little can be concluded from the data, however it does not appear genetic differences in the serotonergic system are particularly important.
And even though some studies have proposed a link between specific serotonin-related genetic polymorphisms and withdrawal, there is little evidence backing up these associations. Murata (2010) reported the C(-1019)G polymorphism in the 5-HT1A receptor gene may associated with paroxetine withdrawal effects. This was taken to indicate at least some aspects of withdrawal may be related to hyperserotonergic activity, hypothetically from unusually low autoreceptor activity and therefore less inhibition of serotonin release, leading to effects on postsynaptic receptors like 5-HT1A and 5-HT2C that contribute to withdrawal (Taylor, 2007 ; Stahl, 2013). In Europeans, no significant evidence was found for a role of 5-HTTLPR genotype (affecting SERT activity) in discontinuation due to adverse effects (Crawford, 2013) although because the number of discontinuation events was low, the authors could not rule out a clinically important effect of genotype, with the S allele potentially associated with a higher risk of discontinuation. And in East Asians there was some evidence the S allele reduces discontinuation risk (Crawford, 2013).
Chronic exposure to SSRIs significantly reduced the mRNA level of 5-HT1B receptors, though this was only seen with paroxetine and fluoxetine, not sertraline. After 3-14 days of withdrawal, mRNA levels were no longer different from the control values.
Chronic TCA (imipramine, amitriptyline, iprindole, and clomipramine) or fluoxetine was tested against metergoline in rats. No antidepressant significantly changed the binding of serotonin to frontal cortex membranes and no drug significantly changed the serotonin concentration in a membrane preparation from rats. Fluoxetine, clomipramine, and iprindole significantly reduced binding by spiperone, which labels a range of serotonin, norepinephrine, and dopamine binding sites. In rats, chronic clomipramine, fluoxetine, and imipramine significantly reduced the behavioral response to the 5-HT2 agonist 5-MeO-DMT. Animals tested 3 days into antidepressant withdrawal had a reduced response to 5-MeO-DMT if they had been on fluoxetine or clomipramine, while iprindole-treated rats showed o change, and those given imipramine had an enhanced response, like has previously been documented with metergoline and amitriptyline.
The effect of 5-HTP in rats is significantly increased following 4 or 10-day exposure to mianserin, danitracen, or amitriptyline (Mogilnicka, 1979). The authors hypothesized the greater response during acute withdrawal could be related to the prolonged presence of the drug in the brain post-withdrawal.
Chronic fluoxetine causes a drop in striatal dopamine that persists for up to two weeks after discontinuation (Gardier, 1994). Though serotonergic drugs can reduce dopamine release, they may also increase some dopaminergic activity by raising the sensitivity of dopamine receptors in the mesolimbic region, which has been put forth as an explanation for the greater motor response to dopaminergic drugs seen after 2-3 weeks of antidepressant exposure in animals (Collu, 1997 ; Serra, 1979 ; D’Aquilla, 2003).
Either a potentiation of or reduction in various types of dopaminergic activity may occur during withdrawal.
With antidepressants that significantly affect norepinephrine, such as TCAs and SNRIs, prolonged exposure can cause adaptations in the adrenergic system. One of those changes may be a decrease in α2 autoreceptor sensitivity such than during withdrawal there is initially a drop in norepinephrine, but this could then lead to increase norepinephrine release and activity due to one of the normal breaks on noradrenergic activity (α2 autoreceptors) being inhibited.
Imipramine administered for three weeks in rats significantly decreased their stress response to forced swimming, which was associated with a decline in nitric oxide synthase (NOS) activity in the hippocampus, which could reduce nitric oxide activity (Harvey, 2006). Further, withdrawal yielded a significant rise in NOS activity compared to control rats or those still on imipramine and the rats exhibited a depressive-like response during forced swimming. Because the 5-HT2A/5-HT2C antagonist ritanserin attenuated the depressive behavioral effects and reversed the increase in NOS activity, it is possible excess 5-HT2A or 5-HT2C activity contributes to some aspects of withdrawal via an effect on nitric oxide.
Wegener (2003) noted various antidepressant types inhibit NOS, and withdrawal symptoms in animals correlate with increased nitric oxide signaling (Tagliaferro, 2001).
Activation of systems mediating a response to stress seems to exist during antidepressant withdrawal. Michelson (2000) demonstrated antidepressant withdrawal, especially from shorter-acting drugs, led to increased stress activity, with a seemingly associated increase in heart rate.
Preclinical and clinical studies have supported the existence of increased glutamatergic transmission after stress and as an important part of depression (Wegener, 2010 ; Gao, 2011).
Excitatory activity from glutamate can be reduced using NMDAR antagonists like MK-801 and MK-801 has been shown to reverse the behavioral and neurochemical effects of withdrawal from the TCA imipramine (Harvey, 2002).
In seven patients given amitriptyline, desiprameine, or imipramine, turnover of norepinephrine (as shown by an increase in MHPG, a norepinephrine metabolite) was increased and peaked during the second week of withdrawal, though without any associated change in blood pressure or heart rate (Charney, 1982). Two of the patients had a rise in anxiety.
Harvey (2002) discussed an animal study in which acute imipramine withdrawal was associated with a loss of situational stress responsiveness and those behavioral alterations were associated with increased hippocampal glutamatergic NMDAR density and a smaller, perhaps reactive, increase in hippocampal GABA levels.
Because TCAs tend to have a high affinity for muscarinic acetylcholine receptors (mAChRs), where they are antagonists, stopping their use appears to produce a hypercholinergic state. This could account for the common occurrence of vomiting, diarrhea, abdominal pain, and sweating seen when stopping TCA use, since muscarinic receptors control aspects of intestinal and sweat gland function.
The order of affinity among TCAs for this receptor is as follows: protriptyline > amitriptyline > imipramine > nortriptyline > desipramine (Tollefson, 1982). Since the affinity of amitriptyline for central muscarinic receptors is 1/20 that of the protypical anticholinergic atropine (Snyder, 1977) and it is given at a 100x higher dose relative to atropine, it is very likely anticholinergic effects are important for drugs like protriptyline and amitriptyline, whereas because desipramine’s affinity is only 1/340 that of atropine, cholinergic effects could be less important.
In rats, long-term exposure to amitriptyline, desipramine, or amoxapine causes increased sensitivity to the hypothermic effect of oxotremorine, a muscarinic receptor agonist, and withdrawal of those TCAs produces a drop in temperature that is blocked by the muscarinic antagonist scopolamine (Dilsaver, 1987). Mice given amitriptyline for three weeks also show an increase in mAChR density in the forebrain (Rehavi, 1980).
Trials from Dilsaver (1983) found TCA withdrawal could be treated with anticholinergics like atropine.
Also supporting a role of acetylcholine in TCA withdrawal, cholinergic drugs cause many of the same effects as TCA withdrawal, such as diarrhea, nausea, and vomiting. Further, they induce REM sleep, an effect associated with vivid dreams, one of the withdrawal symptoms. Increased acetylcholine activity caused by exposure to acetylcholinesterase inhibitors like diisopropyl fluorophosphate causes GI distress in psychiatric patients as well as depression, irritability, emotional lability, and insomnia in healthy people (Rowntree, 1950).
When the acetylcholinesterase inhibitor EA-1701 was given to 93 healthy volunteers, those who still had enzyme activity greater than 80% showed essentially no symptoms, yet those with a reduction to 41-80% activity had sleep issues (e.g. vivid, uncomfortable dreams and insomnia) and those with the least enzyme activity (10-40%) had depression, fatigue, weakness, GI issues, jitteriness, and tension (Bowers, 1964). Many of the symptoms triggered by cholinergic drugs match the symptoms of TCA withdrawal.
Due to the role of cholinergic activity in withdrawal, symptomatic relief has sometimes been reported upon giving anticholinergics like atropine or benztropine.
Treatment and Prevention
The primary preventative step is to slowly taper an antidepressant for months, though this is unnecessary if a drug has only been given for 4-6 weeks. Some authors have argued a year of dose reduction could be desirable for minimizing withdrawal symptoms, but the longest taper periods may not be easy or practical to implement. Tapering studies have produced mixed results in general, probably attributable to the taper periods being too short.
Tapering for up to two weeks offered little to no preventative effect compared to abrupt withdrawal (Tint, 2008 ; Baldwin, 2006) and a study comparing a 14-day taper to a 3-day taper among SSRI and venlafaxine patients showed no significant benefit to the longer taper (Tint, 2007). Unfortunately, medical guidelines from organizations like the National Institute for Health and Care Excellence (NICE, 2009) and UpToDate (Hirsch, 2019) still recommend 2-4 week tapers.
In an observational study of 1200 patients given tapering strips to assist with discontinuation, Groot (2018) found 71% were successful when using the strips to taper over a median of 56 days. The top medications among these patients were paroxetine (47%) and venlafaxine (43%). Importantly, 62% had previously been unsuccessful in at least one withdrawal attempt.
When the withdrawal period is stretched out to multiple months, outcomes are generally pretty good. A study using a 4-month SSRI taper showed patients had half as many symptoms as those discontinuing abruptly (van Geffen, 2005) and a study of people discontinuing paroxetine over a mean of 39 weeks found the taper group had a 6% incidence of withdrawal compared to a 78% incidence with abrupt discontinuation (Murata, 2010).
Linear dose reductions are unlikely to be optimal despite usually being recommended. As discussed by Horowitz (2019), linear reductions cause progressively worse withdrawal symptoms because the decline in SERT occupancy disproportionately increases (i.e. a linear dose change causes a nonlinear activity change). Exemplifying this, Horowitz notes the estimated drop in SERT inhibition will only be 3% when reducing citalopram by 5 mg from 20 mg to 15 mg, yet the final 5 mg reduction from 5 mg to 0 mg will cause a sudden 58% reduction in inhibition.
As a solution, it is recommended the dose be reduced by an amount corresponding to a certain level of SERT occupancy change rather than a certain milligram dose change so that each step down is relatively similar. Essentially, a hyperbolic change in dosing will enable a linear change in activity, which is preferable.
Extended taper schedules are not practical when a patient needs to switch medications, so an abrupt switch or a cross-taper will be useful so long as the two drugs do not dangerously interact and the prescriber has enough expertise to carefully initiate the switch (Tint, 2008). Otherwise, a washout period of a couple days can be used (shorter than the usual 7+ day washout) to reduce withdrawal effects while still offering a smaller interaction risk than an abrupt switch or cross-taper (Keks, 2016).
Baldessarini (2010) found gradual tapers of 2 weeks or longer were superior to 1-7 day tapers among 398 patients given an antidepressant for a mean of 8.5-months. Those discontinuing rapidly had a crude median time to first new depression or panic episode that was shorter by 4.8 months. The impact of gradual tapers on latency to new illness was greater with short-acting drugs relative to longer-acting medications.
Why People Discontinue
There are many reasons someone may decide to stop taking their antidepressant. The top reasons are common side effects and a lack of efficacy, but people also cease treatment due to believing they are no longer at-risk of depression, experiencing unusual and severe side effects, or being worried about the safety of their prescribed drug. Long-term use can cause GI symptoms, weight gain, and sexual dysfunction, all producing a desire to stop treatment.
An interview of 87 patients given SSRIs found a major reason for discontinuation was uncertainty about the benefits of SSRIs in general and/or about their continued necessity (Leydon, 2007). Barriers to discontinuation in the same group were a fear of withdrawal and uncertainty as to what they would be like without medication, even if they had not received a clear benefit from treatment. When a physician believes treatment can be stopped, it is therefore important for patients to be reassured about the low chance of tapering failing to be adequate for minimizing withdrawal.
Sometimes patients have a good reason to stop. A lack of efficacy after multiple months of treatment means the drug is unlikely to be suitable, so it is best to cease treatment and switch to a new therapy. And if someone is experiencing paradoxical negative effects like an increase in depressive symptoms, the treatment should be switched.
When antidepressants are prescribed for panic disorder there is a risk they will trigger depression, an issue that has been reported for decades. Depression sometimes appears in patients given TCAs, SSRIs, and other antidepressants for panic disorder, even in those with no history of mood disorder (Aronson, 1989 ; Fux, 1993 ; Brown, 1995 ; Otto, 1996 ; Fava, 2001, Noyes, 1989). A study of 80 patients administered fluvoxamine for panic disorder reported 7/80 developed depression and their symptoms were alleviated when the drug was stopped and a TCA or clonazepam was given instead (Fux, 1993). Symptoms then reappeared when fluoxetine was given, suggesting some panic patients are sensitive to the effects of serotonergic drugs, while others could be sensitive to all monoaminergic antidepressants.
Among depressed patients, targeting has historically been somewhat poor, with physicians sometimes prescribing for mild symptoms that do not require medication while failing to treat more severe cases (Kendrick, 2001 ; Kendrick, 2005). Also, the medications may be prescribed to people who are only experiencing temporary stress or low mood, which do not call for long-term treatment. This means a fair number of people are on these medications despite not needing them, a notion supported by (Cruickshank, 2008) research suggesting a third to half of long-term antidepressant users do not have an evidence-based reason to continue treatment and could therefore attempt discontinuation. However, it should be noted research on multi-year antidepressant use, much less multi-decade, is sorely lacking, making it difficult to say what the best course of continued care is for long-term users since there is very little evidence to base decision making on.
There is a recognized risk of mania or hypomania during antidepressant use, either from worsening of symptoms among bipolar patients (El-Mallakh, 2015 ; Fava, 2003) or from the appearance of new symptoms among unipolar depressed patients. If either of these occur that is also a fair reason to discontinue the antidepressant.
Stopping at the Right Time
Figuring out when to end treatment is difficult because no reliable predictors of post-withdrawal outcome exist (Berwian, 2017). Some predictors may end up being supported, such as superior indicators of genuine treatment response or a patient’s number of previous depressive episodes, but nothing can currently be used to consistently predict if a patient will do fine off their medication. Between this and research showing continued treatment reduces the risk of relapse, it can be useful to continue therapy as long as it is tolerated.
A review of 27 studies (3037 depressed patients) looked at the outcome of continued treatment vs. discontinuation (Viguera, 1998). 17 studies used a TCA, 5 used an MAOI, and 5 used an SRI. Patients in the discontinuation group had been stabilized on their medication for a mean of 5.8 months after recovery from their depressive episode. The relapse rate (% with a return to depression) was 6.24% per month among discontinued patients compared to 1.85% in those continuing treatment.
Viguera (1998) also showed the estimated time to a 50% risk of relapse was 3.37-fold greater in people still on an antidepressant (48 months vs. 14 months). Length of treatment and the speed of withdrawal had little effect on relapse risk, but a patient’s number of previous depressive episodes was an influential factor, with patients who had 3+ prior episodes showing a 71.7% survival rate at 2-years when on medication compared to 14.7% without medication (a 4.88-fold difference), whereas patients with only 0 to 1 prior episodes had a 2-year survival rate of 64.1% with treatment and 52.7% without, a nonsignficant difference. This could mean people with a less extensive history of depression have a greater chance of doing well without medication.
A meta-analysis of 28 studies (5233 patients) evaluating antidepressant use for anxiety found discontinuation increased the odds of relapse 3.11-fold compared to continuing treatment and the time to relapse was shorter among patients who discontinued (Batelaan, 2017). Though most of the studies had involvement from pharmaceutical companies and the authors of the analysis found six unpublished studies with a different distribution of results (4 negative, 1 positive, and 1 unknown), suggesting there is publication bias in favor of research supporting continued treatment.
Gueorguieva (2017) reported data from four double-blind discontinuation trials of duloxetine and fluoxetine involving 1462 people. Active treatment did reduce the risk of relapse, but the protective effect vs. placebo was only around 13 percentage points (reducing the relapse rate from 46% to 33%), a much smaller effect.
Based on these studies, continuing treatment for depression or anxiety does come with a notable protective effect in terms of relapse risk, but the magnitude of that effect is unknown. It is best to continue treatment when it is tolerated, but withdrawal can be considered once a depressive episode has been successfully treated and a patient has completed 6-9 months of maintenance therapy after recovery (Keks, 2016).
Making Discontinuation Easier
Some patients experience a lot of anxiety when confronted with a discontinuation plan because they are fearful about withdrawal. In these cases, as well as in people who experience emergent anxiety and depression during withdrawal, therapy could be helpful. Cognitive therapies and mindfulness-based therapies appear to reduce relapse risk while also helping people discontinue their medication (Schmidt, 2002 ; Cromarty, 2011 ; Kuyken, 2015).
The UK’s Improving Access to Psychological Therapies (IAPT) program has been shown to help 1 in 6 patients stop their antidepressant by providing supportive psychotherapy (NHS Digital, 2014).
Cognitive behavioral therapy (CBT) has not been extensively researched for helping with antidepressant withdrawal and residual depressive effects, but the available research is positive. Depressive patients successfully treated with antidepressants showed fewer residual symptoms after discontinuation if they also received CBT and they had a 15% relapse rate by 2 years compared to a 35% rate among people who only received normal clinical management after stopping. Because non-drug depression therapy comes with few downsides or risks, it is worth pursuing that form of therapy after completing an antidepressant course rather than stopping all forms of depression treatment.
In a small case series involving 3 patients with paroxetine-induced withdrawal issues, CBT was also helpful, though little can be determined from an uncontrolled review of three cases (Belaise, 2014).
When someone experiences withdrawal symptoms after stopping an antidepressant, the initial response should be to avoid any treatment because of the high likelihood of the symptoms resolving on their own. If the symptoms do not resolve or they increase to an unsustainable level, the same antidepressant should be restarted, followed by a slow taper. While a lower dose of the medication may be effective, the original dose has the greatest chance of working.
If restarting the prior antidepressant does not work, is not possible, or is not warranted, other medications can be used. For TCAs, an anticholinergic like atropine, benztropine, or trihexyphenidyl can be provided for ~1 week to alleviate some of the symptoms (Wolfe, 1997). With SSRIs, SNRIs, and antidepressants generally, benzodiazepines can be helpful for alleviating insomnia, anxiety, muscle tension, and stress during withdrawal. SSRI and SNRI users tend to have an unpredictable response to starting a new antidepressant as a solution to withdrawal. Occasionally a switch is useful, but typically it is best to restart the original drug if an antidepressant still needs to be given.
Wolfe (1997) recommends using a β-blocker, such as a small dose of propranolol, in cases of akathisia (a movement disorder where patients feel restless and cannot stay still) where the antidepressant cannot be restarted.
Antidepressant withdrawal was first described in the 1950s in people receiving imipramine, the first TCA, soon after it entered the market (Mann, 1959 ; Andersen, 1959 ; Kuhn, 1957). Mann (1959) reported half of patients had fatigue and muscle pain during withdrawal.
When SSRIs entered the market they were not associated with withdrawal, but between 1988 and the late 1990s, at least 46 case reports of withdrawal from paroxetine, fluvoxamine, fluoxetine, and sertraline were published (Therrien, 1997). Reactions were reported after as little as 3 weeks of use, but the average duration of use leading to withdrawal was 31 weeks. Symptoms typically lasted no more than a month. Yet even into the 1990s some were arguing antidepressant dependence was less of an issue that it really is, attributing changes like tolerance to a loss of placebo response, the impact of psychosocial factors, or to rapid cycling potentially induced by treatment (Haddad, 1999).
Based on the symptoms reported in published cases of SSRI withdrawal, a 43-symptom checklist was created, known as the Discontinuation Emergent Signs and Symptoms (DESS) checklist (Rosenbaum, 1998). DESS would eventually become the standard checklist to use when investigating antidepressant withdrawal. Based on studies looking at the prescribed doses for SSRIs and TCAs, most prescriptions were sub-therapeutic or at the lower end of the recommended treatment dose range, signs that were taken to indicate relatively little or no tolerance (and likely withdrawal, by extension) occurs with these drugs.