Your new patient is a 53-year-old woman who started venlafaxine 75 mg/day for post-partum depression 19 years ago. Withdrawal symptoms ended a prior attempt to reduce to 37.5 mg/day, suggesting potential problems if she tries again. She is now determined to stop the antidepressant. You agree to help her plan an approach more likely to succeed. During a 30-min follow-up appointment tomorrow, how will you advise her?

Patients and prescribers have a variety of reasons for wanting to stop an antidepressant. Many people start drug therapy at a stressful period in life but adjust over time to a loss or develop non-drug coping skills. Medication may become superfluous.¹ For up to 30% – 50% of people taking antidepressants, no evidence-based reason is found to continue.^2,3,4,5,6 Without compelling clinical reasons to continue an antidepressant, it may simply be time to attempt a return to life without medication.

Women are especially likely to receive long-term prescriptions. For example, an Australian longitudinal study on younger women’s health from mid-2012 to 2019 found a mean duration of drug treatment of 2.4 years in women who refilled an initial prescription (N = 4416 born after 1945).⁷ In contrast, most guidelines recommend 6–12 months of treatment for an episode of anxiety or depression, reserving longer-term treatment for severe and recurrent cases. However, most people who are prescribed antidepressants have mild conditions.^8,9

When patients treated with antidepressant drugs improve, it is natural human psychology – but faulty logic – to attribute this with certainty to a pharmacological effect. Most people improve with time alone.¹⁰ What appears to be obvious clinical logic should be tempered by understanding that in randomised clinical trials (RCTs), the average difference between antidepressants and placebo is about 2 points on the 52-point Hamilton Depression Rating Scale. And the proportion of participants in RCTs who achieve clinically meaningful improvement – compared with placebo – varies from a high of about 15% to a low of about 10%.¹¹

This translates to a number needed to treat for clinical benefit (NNT) of about 7–10.¹² The large majority would do equally well without an active drug. Even when drug therapy at first seems clearly beneficial, pharmacological tolerance may reduce the effect over time, an inevitable consequence of the brain’s adaptation.¹³

For many people, adverse effects of antidepressants outweigh benefits. For example, 50% experience treatment-emergent sexual problems with arousal, erectile dysfunction and orgasm.¹⁴ Emotional numbing, cognitive and memory impairment and sleep disturbance are also common.¹⁵ Evidence of analogous effects in healthy volunteers taking antidepressants indicates that such adverse effects do not arise from a mental illness alone.^16,17

Other harms are not clinically obvious. Antidepressant RCTs are typically small, brief and enrol relatively healthy participants; longer-term harms in the real world are detectable only from pharmacoepidemiologic studies. While epidemiological associations must account for confounding factors and do not prove causation,¹⁸ antidepressants (including selective serotonin reuptake inhibitors [SSRIs] and venlafaxine) have been associated with increased falls, fractures, osteoporosis, bleeding risk, strokes, dementia, cardiovascular disease, weight gain, cataracts and premature mortality, especially in older people.^{19,20,21,22,23} An increased risk of withdrawal after long-term use may be another unanticipated harm.²⁴

Few RCTs have studied how best to stop antidepressants, mostly in people treated for recurrent depression by specialists (total N = 4995 people).²⁵ At least two deprescribing trials are now underway in Australia.^26,27,28 Several cohort studies have compared traditional linear tapers and gradual hyperbolic tapers (described below).²⁹

Current recommendations, guidelines and algorithms are also based on the clinical experiences of clinicians and patients – not experimental results. Just as people differ in their tolerance for drugs and dose increases, their experiences of dose reductions vary. This is similar to other drugs that induce pharmacological tolerance and dependence (e.g. nicotine, alcohol, benzodiazepines and opioids). After antidepressant treatment for more than a few weeks or months, many patients may require a relatively cautious approach to deprescribing.

Unsurprisingly, the more slowly antidepressants are stopped, the lower the risk of withdrawal effects or apparent relapse of depression.³⁰ A meta-regression analysis of antidepressant discontinuation RCTs found no increase in depression relapse when doses were tapered over 6 months compared with continuing drug treatment.³¹ However, fast tapering over 1 month or 2 months was associated with significantly increased relapse in the largest RCT of deprescribing.³² The meta-regression authors emphasise that antidepressant withdrawal can easily be mistaken for a relapse of depression.³¹

The pattern of tapering is also thought to be important. The relationship between the dose and blood concentration of a drug and its effects on receptors in the brain is believed to be hyperbolic. Drug effects rise steeply at doses near the lower end of the approved dose range. But at higher doses and concentrations, many drug effects plateau.³⁰ Hyperbolic dose tapering attempts to mimic this relationship inversely. As the daily dose declines, reductions become progressively smaller. This can be approximated by reducing at an arbitrary percentage, such as 10% per month of the most recent daily dose.

Pragmatic alternatives recommended by many guidelines include stepwise dose reductions over a few weeks, a ‘slow slide’ or substitution of the long half-life SSRI fluoxetine for related antidepressants.³³ However, even the long elimination half-life of the active metabolite norfluoxetine (mean 9 days, range 4–16 days)³⁴ does not avoid withdrawal in about half of people who take fluoxetine – in whom withdrawal may be delayed for 4–6 weeks after stopping.³⁵

Some people need a very gentle approach. In the United Kingdom (UK), a cluster RCT of general practices found that traditional stepwise tapering (e.g. sertraline 200 mg/day, 150 mg/day, 100 mg/day, 50 mg/day, 25 mg/day, stop) was unsuccessful for 58% of patients who had taken at least 9 months of antidepressant therapy for depression or anxiety (N = 274 assessable at 6 months).³⁶

To reduce withdrawal symptoms markedly, a cautious approach to tapering antidepressants (including hyperbolic dose reductions) is now recommended by the UK National Institute for Health and Care Excellence (NICE) and a number of UK specialty organisations.^37,38

The rate and pattern of tapering are clinical decisions specific to the circumstances and needs of individual patients. The Canadian Medication Appropriateness and Deprescribing Network provides a convenient, concise and lucid guide for patients, their family or friends and clinicians (also available in French).³³

As a general rule, people who have taken an antidepressant for more than 4 weeks should preferably not stop abruptly. A practical approach to avoid or minimise withdrawal is a test dose reduction, using the patient’s response to determine further steps. Barring an urgent need to stop the antidepressant, slow decrements are reasonable. One can accelerate the pace as tolerated. A very slow taper may imply a few weeks or months of unnecessary drug exposure (at a very low dose). On the other hand, excessively rapid dose reduction potentially risks akathisia, suicide or possible years of disability. Poor outcomes are not always reversible.¹

Patients can be risk-stratified based on the duration of use, type of antidepressant, past experience of stopping and dose.^39,40 People who have taken an antidepressant for only several weeks are at low risk of serious withdrawal (but not no risk), compared with those with years of exposure. Past experience of withdrawal and high doses also predict the future risk. But the smallest doses used clinically affect amine transporters and brain receptors almost as much as maximum doses.³⁰ Therefore, even a ‘low’ dose of an antidepressant often confers a substantial risk of withdrawal upon discontinuation.

Without direct comparisons between antidepressants after long-term use, withdrawal risk categories are inherently arbitrary. Table 1, adapted from a review of data available in 2022,³⁵ shows categorical risk estimates for frequent or severe withdrawal and published incidence estimates for antidepressants commonly used in Canada.

Two analyses of the World Health Organization’s adverse drug reaction (ADR) database also identified paroxetine, venlafaxine or desvenlafaxine, duloxetine, sertraline and citalopram or escitalopram as the highest risk for withdrawal.^41,42 However, this could reflect ADR reporting biases for newer antidepressants. A 2023 Canadian guideline on the treatment of depression provides different ratings but does not cite sources.⁴³ And a 2024 systematic review of short RCTs and observational studies also identified imipramine (the original tricyclic antidepressant) as especially problematic.⁴⁴

For low-risk patients (weeks of use, low-risk antidepressant), a rule of thumb is to start with a 25% dose reduction. For patients deemed at moderate risk (months of use, moderate-risk antidepressant), reducing by 10% may be more appropriate. For patients expected to be at high risk for withdrawal (years of use, high-risk antidepressant, significant prior problems with withdrawal), a 5% initial reduction is unlikely to cause serious problems.

After starting a dose taper, experts in deprescribing antidepressants recommend maintaining and monitoring patients clinically for 2–4 weeks per step to assess the response. In ‘slow slide’ or ‘hyperbolic’ tapering, milligram reductions in daily dose become smaller and smaller as the dose declines. Table 2 shows a hyperbolic tapering schedule that would last well over 1 year for the Vignette patient who took venlafaxine 75 mg/day for 19 years, derived from the Maudsley Deprescribing Guidelines: Antidepressants, Benzodiazepines, Gabapentinoids and Z-drugs.¹

Schedules recommended by the Canadian Medication Appropriateness and Deprescribing Network are somewhat faster.³³ The ongoing Australian RELEASE deprescribing RCT also provides a series of possible ‘slow’ and ‘slower’ schedules for 15 antidepressants.⁴⁵

English and Dutch physicians with substantial deprescribing experience indicate that people at low risk can generally stop in 6–9 months and people at moderate risk in 9–18 months. But high-risk patients may require 2 years or longer. Other experienced clinicians indicate that people treated only briefly with an antidepressant may be able to stop within 2–4 weeks without problematic withdrawal.³³ As with other habituating drugs like nicotine, alcohol and caffeine, some patients may simply prefer to stop ‘cold turkey’ after short-term use of antidepressants.

Difficult withdrawal symptoms often improve by pausing dose reductions or retreating a step or two in dose reduction and slowing the taper. Profound suffering can necessitate reinstating the original dose. Prescribing other drugs to suppress withdrawal substitutes another potential problem, so it is clinically more sensible to interpret withdrawal symptoms as a useful signal to slow the tapering process.

Vignette resolution: Your new patient has taken venlafaxine 75 mg/day for 19 years. Given her prior experience trying to reduce the dose to 37.5 mg/day, she is at high risk for a withdrawal syndrome if she stops precipitately. But you agree that stopping venlafaxine is reasonable and propose a cautious approach. After a thorough discussion, you provide her with written information,³³ and plan to request preparation of progressively smaller doses with the help of an experienced pharmacist. At the next visit, you and your patient can agree on a supervised taper and a first follow-up in 2 weeks.