There is strong evidence for the relapsing and recurring nature of major depression in patients presenting to hospital.^1,4,5 Much less is known of the natural and treated history of depressive illness in general practice. Without continuing treatment, about 1 in 3 hospital patients will have a relapse of a recently treated acute episode within 6 months, and up to half will have a relapse or a new episode of depression within 2 years.⁴ Recurrence is more likely if: there is a history of previous episodes; the acute illness is severe; residual symptoms persist at the end of treatment; the patient has ongoing chronic medical or social problems (e.g. unemployment or relational difficulties), lacks social support, or is aged under 25 or over 60 years at the onset of the illness.^2,4,5 Current guidelines for antidepressant drug therapy distinguish between an initial acute treatment phase and a continuation phase, which are recommended for all patients, and a maintenance phase, which is indicated for selected patients with recurrent severe depression.^2,3

The acute phase covers the initial weeks of treatment, until the patient has achieved a significant clinical response, usually with at least 40-60% improvement in symptoms.

About 30% of patients remain depressed despite initial antidepressant drug therapy.⁶ Data from the USA suggest that patients showing no clinical improvement after 4 weeks' treatment with a high dose of an antidepressant do not benefit from continuing the drug, but continuing until 6 weeks is worthwhile in patients showing even modest improvement.⁶ With the rather lower doses used in the UK, a 6-8 week trial of initial treatment is probably justified, increasing the dose in those who show a partial response, before deciding (after reviewing the diagnosis) to change to a different first-line treatment.

The continuation phase aims to consolidate the initial acute response. Placebo-controlled trials in patients attending hospital have shown that treatment lasting until the patient has been fully free of symptoms for 4-6 months roughly halves the rate of relapse.^3,7,8 Provided it is tolerated, continuation treatment should be given in the antidepressant dose that brought about the initial response.^2,5 Based on these hospital data, most experts recommend continuation therapy for general practice patients with depression,^2,3 but comparable efficacy studies for treatment in the community are needed.

The aim of continuation treatment is to achieve at least 4 months of full recovery (with normal mood and functioning equal to that before the illness) before withdrawing treatment.⁵ Incomplete recovery is the most important risk factor for relapse within 1 year of treatment withdrawal.⁷ Before withdrawing therapy, therefore, evidence of depressive symptoms should be sought. If symptoms are still present, treatment should not be stopped. Contributory causes (e.g. medical or social problems, substance abuse, non-adherence to treatment) should be looked for and the treatment regimen may need to be reviewed.

The decision to go on to maintenance therapy, rather than stop treatment at the end of the continuation phase, must be made on clinical grounds, but there is no clearly agreed threshold. A key criterion is the number of previous episodes: it is reasonable to consider maintenance treatment if a patient has had two or more previous severe episodes of depressive illness (i.e. requiring hospital attendance) over the preceding 5 years.³ Such patients should normally be under the care of a psychiatrist.

Hospital-based, placebo-controlled studies have demonstrated efficacy in patients with highly recurrent depression, from maintenance therapy lasting 1-3 years with older antidepressants such as amitriptyline, nortriptyline and phenelzine,^3,9,10 and for 3-5 years with imipramine.^11,12 Provided it is tolerated, such treatment (alone or with psychotherapy) appears optimally effective when given in the dose that achieved a full response in the initial treatment phase.^5,11 Efficacy as maintenance has also been shown for selective serotonin reuptake inhibitors (SSRIs), but few studies have lasted longer than a year.^3,13

We know of no adequate trial of maintenance therapy in patients managed solely in general practice. Although many patients are given antidepressant treatment long term, a recent study in South Wales found that, of those receiving a tricyclic long term (average 3 years), only one-third were taking an adequate therapeutic dose, while two-thirds of all patients on long-term antidepressants still had moderate or severe symptoms of depression.¹⁴

The decision to stop maintenance antidepressant therapy may be straightforward, as when a patient has a serious adverse reaction or when unwanted effects which were tolerated during the acute phase (e.g. weight gain with tricyclics, sexual dysfunction with SSRIs) prove unacceptable during longer-term therapy. Intercurrent illness (e.g. myocardial infarction, cardiac arrhythmia), co-administration of potentially interacting drugs, or pregnancy, may make drug treatment unsafe or unsuitable, and treatment may also have to be stopped or changed if it ceases to be effective. Up to one-third of patients have a return of depression despite long-term antidepressant treatment in full dosage.⁹

The decision is most difficult when a patient with a history of recurrent depression has remained well on maintenance therapy for several years and feels confident to discontinue. The doctor and the patient will need to consider the severity of previous episodes, how straightforward or difficult the episodes were to treat, the effect they had on the patient's personal, family and working life, and whether a future episode might carry a significant risk of suicidal behaviour or risk to others. Careful discussion of these factors with the patient (and preferably with a key family member) is essential when considering whether to stop long-term treatment. If medication is withdrawn, the patient must be carefully followed up: this is the only way of finding out if continued drug treatment is still needed.

Before withdrawing an antidepressant, patients should also be warned of the possibility that stopping treatment can itself lead to a cluster of symptoms which are commonly referred to as a 'discontinuation syndrome'.

Discontinuation syndromes can occur with any antidepressant; they usually start abruptly within a few days of stopping the drug and resolve quickly (usually within 24 hours) if the drug is restarted. In general, it should be possible to distinguish discontinuation syndromes from true relapse of depression, which is uncommon in the first week after stopping treatment and resolves more slowly when the drug is restarted. On the other hand, re-emergent symptoms of depression may be a feature of the discontinuation syndrome itself,¹⁵ so the distinction is not always clear-cut. Ignorance of discontinuation reactions could have led, in the past, to some overestimation of relapse rates when antidepressants were withdrawn.

Discontinuation syndromes on withdrawal of tricyclic antidepressants are now well described.¹⁶ Gastrointestinal (e.g. abdominal pain, nausea, vomiting and diarrhoea) and 'flu-like symptoms, fatigue, anxiety and agitation, nightmares and sleep disturbance are the most common features. Movement disorders, such as akathisia, and behavioural activation with hypomania, are occasional effects.¹⁶ Controlled data on the frequency of discontinuation symptoms are few, but longer therapy and abrupt withdrawal from a high dose appear to be contributory factors. Discontinuation symptoms have been reported in up to 50% of patients suddenly stopping high-dose imipramine,¹⁶ and in 30% of patients undergoing supervised withdrawal of clomipramine.¹⁷

There are now many reports of discontinuation reactions with SSRIs.^13,18 Typically symptoms begin within 24-72 hours of stopping the drug and last 1-2 weeks, but occasionally much longer.¹⁹ The most common symptoms are dizziness, nausea, lethargy and headache; other symptoms are anxiety, paraesthesia, 'shock-like' sensations, balance problems, tremor, sweating, insomnia and nightmares.^17,18

Studies specifically designed to assess discontinuation symptoms with SSRIs commonly report rates of around 20% or more.^18,20 A retrospective study of 171 patients who had undergone supervised withdrawal of clomipramine or an SSRI found that discontinuation symptoms were most common in those stopping clomipramine (31%), and more common after paroxetine (20%) and fluvoxamine (14%) than after sertraline (2%) or fluoxetine (0%).¹⁷ Analysis of Yellow Cards sent to the Committee on Safety of Medicines up until 1994 found that discontinuation reactions formed a greater proportion of all reports received on paroxetine (5.1%) than on fluvoxamine (0.4%), sertraline (0.9%) or fluoxetine (0.06%).¹⁹ Follow-up of 217 consecutive reports of discontinuation reactions with paroxetine found that 58% were considered 'moderately severe' and 21% 'severe'.¹⁹

In a study involving 242 patients, open-label maintenance treatment with paroxetine (20-60mg daily), sertraline (50-150mg daily) or fluoxetine (20-60mg daily) was interrupted for 5-8 days by double-blind placebo substitution, to mimic intermittent non-adherence to treatment.¹⁵ Discontinuation reactions were more common in paroxetine-treated (66%) and sertraline-treated patients (60%) than in the fluoxetine group (14%). The long plasma half-life of fluoxetine (4-6 days) and its active metabolite, norfluoxetine (4-16 days) means that discontinuation symptoms might not emerge during a 1-week interruption. This does not, however, rule out a potential for late effects on treatment cessation. A shorter plasma half-life appears to be an important, but not the sole, determinant of discontinuation reactions.^18,19

Discontinuation symptoms (occasionally severe) have also been reported after ▼venlafaxine cessation. They include headache, dizziness, nausea, diarrhoea and shock-like sensations.¹⁸

Debate about whether the 'discontinuation' syndrome is different from classical 'withdrawal' is probably largely a matter of semantics; the crucial issue is the question of dependence.²⁰ A 'withdrawal syndrome' means the emergence, on stopping a drug, of a well-defined group of symptoms, with a predictable onset, duration, and offset of action, containing psychological and bodily symptoms not previously complained of by the patient.²¹ There is a general perception that withdrawal is indicative of dependence, although as currently defined, dependence requires the presence of other features such as a compulsive desire to take the drug, difficulty controlling level of use despite evidence of harm, neglect of other pleasures, and evidence of tolerance.²² A discontinuation syndrome typically occurs on cessation of a regular therapeutic dose, and is not, on its own, taken to imply dependence, as so defined. From the patient's point of view, however, discontinuation symptoms would be highly important if they resulted in any marked degree of difficulty in stopping the treatment, and might well feel tantamount to dependence.²⁰

Awareness of discontinuation symptoms, and avoiding abrupt withdrawal of any antidepressant (except for a compelling clinical reason) are the most important safeguards in minimising problems. Patients should be advised not to omit doses or stop treatment suddenly, and should be forewarned about possible symptoms when the treatment is discontinued. Gradual tapering of the dose seems prudent: although there are no controlled data, our consultants recommend tapering over 6-8 weeks after a 'standard' 6-8 months' treatment, and even more gradual tapering, say by one-quarter of the treatment dose every 4-6 weeks, after maintenance. After courses lasting less than 8 weeks, discontinuation over 1-2 weeks is considered safe, and minimises delay when a change in medication is needed.

Many discontinuation symptoms are mild and short-lasting, and manageable with reassurance and explanation.¹⁸ In some patients, however, the symptoms are severe and make normal functioning impossible, even with slow withdrawal. In these patients, increasing the dose of antidepressant and then tapering it more slowly may help.¹⁶ Some specialists recommend switching from a drug with a shorter half-life to one with a longer half-life, but we know of no study of this approach.

When withdrawing antidepressant medication, a careful watch should be kept for signs of returning depressive symptoms. The development of even mild symptoms should be taken seriously; if there is clear evidence of impending relapse or recurrence, treatment should be restarted at a dose which was previously effective. A sound relationship between doctor and patient is important in detecting early relapse and can provide valuable support and reassurance during any discontinuation symptoms.

Both in hospitals and in the community, patients with acute depressive illness are now being treated with antidepressants for longer. In both settings, treatment in full therapeutic dose, for 4-6 months after resolution of depressive symptoms, is considered a minimum by most psychiatrists, based on studies in patients attending hospital. Long-term maintenance therapy, also in a full treatment dose (and often with other psychological treatment) can prevent recurrence of depression in hospital patients with severely recurrent illness, but has not been tested in general practice. We see few, if any, sound grounds for starting long-term maintenance except under the direction of a psychiatrist.

Deciding when to withdraw treatment, especially maintenance treatment, can be difficult and requires careful discussion of the potential benefits and risks. If treatment is discontinued, the doctor and the patient need to be alert to the risk of re-emergence of depressive symptoms and to discontinuation effects, which can occur on cessation of any antidepressant. Antidepressants should not be stopped abruptly, nor treatment courses interrupted, unless there is a clear clinical reason, such as a serious adverse effect. We suggest that doctors report any suspected discontinuation reaction to the Committee on Safety of Medicines on a Yellow Card, and certainly any that are severe or make treatment withdrawal difficult.

[M=meta-analysis; R=randomised controlled trial]