11.2.1. Introduction

**Depression is the most common mental health problem of later life affecting approximately 15% of older people (Beekman et al., 1999). Untreated it shortens life and increases healthcare costs, as well as adding to disability from medical illnesses, and is the leading cause of suicide among older people (Lebowitz et al., 1997). Most depression in older adults is treated in primary care (Plummer et al., 1997) but there is evidence of poor detection (Plummer et al., 1997) and sub-optimal treatment (Iliffe et al., 1991). In this population the monitoring of self-harm is particularly important. It is also very important to educate the patient and caregivers about depression and involve them in treatment decisions. Older adults are at risk of co-existing physical disorders, sensory deficits and other disabilities and, therefore, medication needs to be carefully monitored in these groups.

The efficacy of antidepressants in older adults has been summarised in a Cochrane systematic review (Wilson et al., 2001). There is some evidence that older people take longer to recover than younger adults and adverse events need to be carefully monitored for, since they might substantially affect function in a vulnerable individual.

There are a variety of potential differences in older adults in terms of absorption and metabolism of drugs and increased potential for interaction with other drugs. The maxim is, therefore, to start low and increase slowly but it is clear that much more research involving older patients with depression is required on this and other points.

It was possible to review the following pharmacological strategies for the treatment of depression in older adults:

• use of individual antidepressants (amitriptyline, TCAs as a group, SSRIs, phenelzine, mirtazapine, venlafaxine) and St John’s wort; studies were also available for reboxetine but, since this drug is not licensed for the treatment of depression in older adults, it is not reviewed
• augmentation of an antidepressant with lithium
• strategies for relapse prevention.

11.2.2. Use of individual antidepressants in the treatment of depression in older adults

11.2.3. Augmentation of an antidepressant with lithium in older adults

11.2.4. Relapse prevention in older adults

11.2.5. Clinical summary

There is no difference in the efficacy of the various antidepressants for which studies have been undertaken in older adults. There is also no evidence of differences in acceptability. There is no evidence that there is a difference by setting, apart from in primary care, where fewer patients taking paroxetine report adverse events compared with those taking amitriptyline.

With regard to augmenting an antidepressant with lithium, elderly patients appear to be more likely to achieve remission without the addition of lithium. These patients are also less likely to leave treatment early.

It appears to be worthwhile continuing pharmacological treatment in elderly patients with multiple depressive episodes in order to avoid relapse.

These results are similar to those found in the reviews of studies for all adult patients elsewhere in this guideline.**

11.2.6. From evidence to recommendations

The review of pharmacological treatments for older adults was not updated because there were little new data, and the overall conclusions in the previous guideline were that management of older adults should follow general principles. These were based on the fact that older people tend to metabolise drugs more slowly and are more likely to be taking concomitant medication and to be in poorer physical health than younger people. These recommendations are unchanged. However, they have been amended to bring them up to date with current NICE style. Since the publication of the previous guideline, a guideline on the management of dementia has been published (NICE, 2006b). This covers the management of depression comorbid with dementia and, therefore, recommendations relating to this topic have been removed.

11.2.7. Recommendation

11.2.7.1.

When prescribing antidepressants for older people:

• prescribe at an age-appropriate dose taking into account the effect of general physical health and concomitant medication on pharmacokinetics and pharmacodynamics
• carefully monitor for side effects.

11.3.1. Review of the evidence

Although the female preponderance in the prevalence of unipolar depression has been well established (Weissman et al., 1993), relatively little attention has been paid to gender differences in treatment response to antidepressant medication. A meta-analysis of 35 studies published between 1957 and 1991 that reported imipramine response rates separately by sex reported that men responded more favourably to imipramine than women (Hamilton et al., 1996). Some studies since then have suggested that younger women may respond preferentially to SSRIs over noradrenaline reuptake inhibitors (TCAs, maprotiline, reboxetine) with predominantly no difference found for men (Kornstein et al., 2000; Martenyi et al., 2001; Joyce et al., 2002; Baca et al., 2004; Berlanga & Flores-Ramos, 2006). This may be accounted for by a poorer tolerability of TCAs in younger women (Kornstein et al., 2000; Joyce et al., 2002; Baca et al., 2004). Results are inconsistent as to whether men respond better than women to TCAs (Quitkin et al., 2001). A study comparing TCAs and MAOIs found that in patients with atypical depression and associated panic attacks, women showed a more favourable response to MAOIs and men to TCAs (Davidson & Pelton, 1986).

However, the data are not consistent, and several studies have failed to show any significant effect of sex on antidepressant response, for example, when SSRIs were compared with clomipramine in inpatients (Hildebrandt et al., 2003), and no effect of sex has been found with venlafaxine (Hildebrandt et al., 2003), duloxetine (Kornstein et al., 2006), and amfebutamone (bupropion) (Papakostas et al., 2007). A large observational study of sertraline treatment in over 5,000 patients failed to find a clinically relevant effect of sex on response to treatment (Thiels et al., 2005).

Taken as a whole, no convincing data showing differential benefits for antidepres-sants based on sex have accrued since the previous guideline; the GDG considered that the previous recommendations should be removed from the guideline update. However, recommendations from the guideline Antenatal and Postnatal Mental Health: Clinical Management and Service Guidance (NICE, 2007e) should be considered when treating women of childbearing age who have depression.

11.3.2. Recommendation

11.3.2.1.

Do not routinely vary the treatment strategies for depression described in this guideline either by depression subtype (for example, atypical depression or seasonal depression) or by personal characteristics (for example, sex or ethnicity) as there is no convincing evidence to support such action.

11.4.1. Introduction

**Major depression with psychotic features is a disorder with considerable morbidity and mortality. In an epidemiologic catchment area study (Johnson et al., 1991), 14.7% of patients who met the criteria for major depression had a history of psychotic features. The prevalence is higher in samples of elderly patients. The disorder is often not diagnosed accurately because the psychosis may be subtle, intermittent or concealed. There has been a long-standing debate as to whether major depression with psychotic features is a distinct syndrome or represents a more severe depressive subtype. The weight of evidence suggests that severity alone does not account for the differences in symptoms, biological features and treatment response (Rothschild, 2003). The systematic study of major depression with psychotic features has been limited by the fact that the disorder does not exist as a distinct diagnostic subtype in DSM–IV and because of the difficulties in enrolling such patients in research studies. As a result there are few controlled studies on the acute treatment of psychotic depression and no long-term maintenance studies. There is some evidence that patients with major depression with psychotic features exhibit more frequent relapses or recurrences than patients with non-psychotic depression; however, not all studies are in agreement (see Rothschild, 2003). Patients with major depression with psychotic features demonstrate more severe psychomotor disturbance more frequently than patients without psychosis.

11.4.2. Studies considered for review^163,164

Twenty studies were found in a search of electronic databases, six of which met the inclusion criteria set by the GDG (Anton1990, Bellini1994, Mulsant2001, Spiker1985, Zanardi1996, Zanardi2000) and 14 of which did not, mainly because too many participants had been diagnosed with bipolar depression and, therefore, fell outside the inclusion criteria set by the GDG.

Four studies (Anton1990, Bellini1994, Mulsant2001, Spiker1985) looked at augmenting an antidepressant with an antipsychotic and two (Zanardi1996, Zanardi2000) compared a single antidepressant with another. The following comparisons were possible:

• amitriptyline plus perphenazine versus amoxapine
• nortriptyline plus perphenazine versus nortriptyline plus placebo
• amitriptyline plus perphenazine versus amitriptyline
• desipramine plus haloperidol versus desipramine plus placebo¹⁶⁵
• fluvoxamine plus haloperidol versus fluvoxamine plus placebo¹⁶⁶
• paroxetine versus sertraline
• fluvoxamine versus venlafaxine.

In comparisons involving antipsychotic augmentation, efficacy data were available from up to 103 participants and tolerability data from up to 87 participants. In comparisons comparing single antidepressants, both efficacy and tolerability data were available from up to 60 participants. All included studies were published between 1985 and 2001 and were between 4 days and 16 weeks (mean = 7.17 weeks).

All studies were of inpatients, and in one all patients were at least 50 years of age (mean = 71 years) (Mulsant2001). Participants had a diagnosis of major depressive disorder with psychotic features. In two studies (Anton1990, Zanardi2000) up to 25% (the limit allowed in the inclusion criteria set by the GDG is 15%) of participants were diagnosed with bipolar disorder. Two sets of analyses were performed including and excluding these two studies. There was no difference in results, so statements from the analysis excluding these studies are presented below.

11.4.3. Clinical evidence statements¹⁶⁷

11.4.4. Clinical summary

There is no good quality evidence for pharmacological treatments of psychotic depression. However, there are practical problems in recruiting sufficient numbers of patients with psychotic depression and, therefore, practitioners may wish to consider lower levels of evidence.**

11.4.5. Recommendation

11.4.5.1.

For people who have depression with psychotic symptoms, consider augmenting the current treatment plan with antipsychotic medication (although the optimum dose and duration of treatment are unknown)¹⁶⁸.

11.5.1. Introduction

**Depression with atypical features is described in DSM–IV (APA, 1994). The introduction of a formally defined type of depression with atypical features was in response to research and clinical data indicating that patients with atypical depression have specific characteristics. The classical atypical features are over-eating and oversleeping (sometimes referred to as reverse vegetative symptoms). The syndrome is also associated with mood reactivity, leaden paralysis and a long-standing pattern of interpersonal rejection sensitivity. In comparison with major depressive disorder without atypical features, patients with atypical depression are more often female, have a younger age of onset and a more severe degree of psychomotor slowing. Coexisting diagnoses of panic disorder, substance misuse and somatisation disorder are common. The high incidence and severity of anxiety symptoms in these patients increases the likelihood of their being misclassified as having an anxiety disorder. The major treatment implication of atypical depression is that patients are said to be more likely to respond to MAOIs than TCAs. However, the significance of atypical features remains controversial as does the preferential treatment response to MAOIs. The absence of specific diagnostic criteria has limited the ability to assess the aetiology, prevalence and validity of the condition.

11.5.2. Studies considered^169,170

This section brings together studies from other reviews undertaken for this guideline where participants were diagnosed with atypical depression. A separate systematic search of the literature was not undertaken and, therefore, studies of atypical depression using drugs not reviewed for this guideline are not included.**

No new studies were found in the update search for the guideline update.

**In all, three studies from other reviews were of atypical depression (Mcgrath00, Pande1996, Quitkin1990). Two came from the review of phenelzine and one from the review of SSRIs. Data were available to look at the efficacy of phenelzine compared with imipramine/desipramine or with fluoxetine, and fluoxetine compared with imipramine. But there was only tolerability data available for phenelzine compared with fluoxetine. Efficacy data were available from up to 334 patients, and tolerability data from up to 40 patients. All included studies were published between 1990 and 2000. Two were classified outpatient studies and in the other it was not possible to determine the source.

11.5.3. Clinical evidence statements¹⁷¹

11.5.4. Clinical summary

In patients with atypical depression there is some evidence suggesting a clinical advantage for phenelzine over TCAs (imipramine/desipramine) in terms of achieving remission and response. However, compared with SSRIs (fluoxetine), there is evidence of no difference on mean endpoint scores, and insufficient evidence on other outcome measures. There is insufficient evidence for the acceptability and tolerability of any antidepressant.**

11.5.5. From evidence to recommendations

The previous guideline recommended treatment with an SSRI for people with atypical depression. Since this is the treatment of choice for all people with depression, the guideline group decided to remove the recommendation from the updated guideline. They also considered that the other recommendations for treating atypical depression were adequately covered elsewhere in the guideline (cautions about the use of phenelzine, and referring to a mental health specialist), and that no special management of people with atypical depression could be recommended.

11.5.6. Recommendation

11.5.6.1.

See recommendation 11.3.2.1.

11.6.1. Introduction

The term seasonal affective disorder (SAD), introduced by Rosenthal and colleagues (1984) to describe recurrent depressions that have a seasonal pattern and occur annually at the same time each year, includes bipolar depression but most people affected have recurrent unipolar depression (70 to 80%). Winter depression with a seasona pattern is far more common than summer depression with a seasonal pattern. DSM–IV includes criteria for a seasonal pattern for depressive episodes whereas only provisional criteria are given in the research version of ICD–10. The characteristic quality of major depression with a seasonal pattern is that symptoms usually present during the winter and remit in the spring. The symptoms of depression with a seasonal pattern do not clearly delineate it from other types of depression but in reported samples decreased activity was nearly always present and atypical depressive symptoms were common, particularly increased sleep, weight gain and carbohydrate craving.

Depression with a seasonal pattern as a separate diagnosis has been less accepted in Europe than North America, and an alternative view is that major depression with a seasonal pattern is an extreme form of a dimensional ‘seasonality trait’ rather than a specific diagnosis with so-called ‘subsyndromal major depression with a seasonal pattern’ appearing to be common. Nevertheless there are some patients with recurrent major depression who experience a seasonal pattern to their illness, at least for a time. There also appear to be people who experience seasonal fluctuations in mood that do not reach criteria for major depression.

The hypothesis that light therapy (that is, increasing the amount or duration of light exposure) might be an effective treatment is based on the presumption that depression with a seasonal pattern is caused by a lack of light in the winter months. There have subsequently been a number of controlled studies and meta-analyses (for example, Golden et al., 2005) that have concluded that light therapy may be effective. There has been little research into other treatments in patients with depression with a seasonal pattern.

11.6.2. Databases searched and the inclusion/exclusion criteria

Information about the databases searched for published trials and the inclusion/exclusion criteria used are presented in Table 91. Details of the search strings used are in Appendix 8.

Table 91

Databases searched and inclusion/exclusion criteria for clinical effectiveness of psychological treatments.

11.6.3. Light therapy for depression with a seasonal pattern

Depression with a seasonal pattern was not included in the scope of the previous guideline. Light therapy, which has been developed as a treatment specifically for major depression with a seasonal pattern, was therefore not reviewed, but has been included here as an additional review for the guideline update. For this review both published and unpublished RCTs investigating light therapy in patients diagnosed with major or subsyndromal major depression with a seasonal pattern were sought. There are a range of methods for administering light therapy; this review included a range of light treatments such as a light box, light room or visor and dawn simulation. Trials comparing a light treatment with a control condition, another light treatment or light administered at different times of day were included in this review.

A special adviser was consulted regarding a number of issues for this review (see Appendix 3). He advised the GDG that 5,000 lux hours¹⁷² per day is a reasonable minimum dose for light box treatment, but that a minimum effective dose of light administered by a light visor has not yet been established. For the control light condition a placebo light of not more than 300 lux is appropriate. He suggested that a minimum trial duration of a week would be reasonable for evaluating the efficacy of light treatment. His advice was also sought regarding dawn simulation; he suggested that it would be informative to include this type of light treatment in the review and that a simulation of around an hour and a half peaking at 250 lux is an appropriate minimum, with a control condition of a light of less than 2 lux.

Studies considered¹⁷³

In total, 61 trials were found from searches of electronic databases. Of these, 19 were included and 42 were excluded. The most common reasons for exclusion were that papers were not RCTs or participants did not have a diagnosis of depression or subsyndromal depressive symptoms with a seasonal pattern. In addition, studies that used a cross-over design (where participants serve as their own controls by receiving both treatments) were not used unless pre-crossover data were available.

The studies that were found by the search and included in this review varied considerably in methodology. The intensity and duration of light, time of day, mode of administration of light, and the comparison conditions were different across studies. A range of outcomes were reported by the included studies, including the HRSD (termed ‘typical’ depression rating scale to distinguish it from scales measuring depression with seasonal pattern symptoms), and scales adapted for measuring symptoms in depression with a seasonal pattern. These included the Structured Interview Guide for the Hamilton Depression Rating Scale (SIGH) for major depression with a seasonal pattern (Williams et al., 1988), which combines the HRSD with an additional eight items relevant to depression with a seasonal pattern. Some studies report the eight additional items separately. Both typical and atypical symptoms were measured using clinician- and self-rated scales. All data were extracted and can be seen in the full evidence profiles and forest plots (Appendix 16c and Appendix 19c, respectively). Only data for the SIGH for major depression with a seasonal pattern (clinician- and self-rated) are presented here.

Data were available to compare light therapy with a range of control conditions including waitlist, attentional controls and active treatment controls. In addition administration of light in the morning versus evening was compared and dawn simulation was compared with attentional control and with bright light. One study included a combination treatment of light and CBT and one trial reported on light therapy for relapse prevention.

Summary study characteristics of the included studies are presented in Table 92 and Table 93 with full details in Appendix 17c, which also includes details of excluded studies.

Table 92

Summary study characteristics of light therapy studies versus control and morning light versus afternoon/evening light.

Table 93

Summary study characteristics of dawn simulation and relapse prevention studies.

Clinical evidence

Bright light versus waitlist or attentional control

Compared with waitlist control, bright light (either light room or light box) shows a strong effect on symptoms in depression with a seasonal pattern although there are few studies. Compared with attentional controls, such as deactivated negative ion generator, dim red light, and sham light boxes, bright light (either via light box or light visor) shows a small effect on symptoms in depression with a seasonal pattern that was not clinically important. Evidence from the important outcomes and overall quality of evidence are presented in Table 94. The full evidence profiles and associated forest plots can be found in Appendix 16c and Appendix 19c, respectively.

Table 94

Summary evidence profile for bright light versus waitlist or attentional controls.

Bright light versus active treatment control

There were data to compare light therapy with group CBT, light therapy plus CBT, and dim light plus fluoxetine. There was also a study comparing light therapy plus St John’s wort with dim light plus St John’s wort.

Compared with group CBT (tailored to depression with a seasonal pattern) bright light therapy was no better in terms of reducing depressive symptoms in depression with a seasonal pattern, although the effect size is not statistically significant and was graded low quality. However, more participants achieved remission with bright light therapy than with group CBT (52% compared with 37.5%), although the result is not clinically important. Similarly, light therapy appeared to be more acceptable than group CBT with fewer people leaving treatment early (8% compared with 16.7%) although the effect size is not statistically significant. Treatment lasted for 6 weeks.

Combination treatment (bright light plus CBT) was more effective than light therapy alone on both the SIGH for major depression with a seasonal pattern and the BDI, although the effect sizes were not statistically significant. Roughly equal numbers of participants left treatment early.

There appeared to be little difference between bright light therapy and fluoxetine (20 mg) on efficacy outcomes (both treatments given with a sham treatment mimicking the other). Treatment lasted for 8 weeks.

There was no evidence for the efficacy of light therapy combined with St John’s wort compared with a sham light condition plus St John’s wort. There was only a single small 4-week study (n = 20).

Evidence from the important outcomes and overall quality of evidence are presented in Table 95. The full evidence profiles and associated forest plots can be found in Appendix 16c and Appendix 19c, respectively.

Table 95

Summary evidence profile for bright light versus active treatment control.

Morning light versus afternoon/evening light

Three studies compared light therapy administered in the morning compared with light therapy in the afternoon or evening, one of which was in participants with subsyndromal major depression with a seasonal pattern. There were no significant differences in outcome measures for those given light therapy in the morning compared with those given light therapy in the afternoon or evening. Evidence from the important outcomes and overall quality of evidence are presented in Table 96. The full evidence profiles and associated forest plots can be found in Appendix 16c and Appendix 19c, respectively.

Table 96

Summary evidence profile for morning light versus evening light.

Dawn simulation versus attentional control or light therapy

Three studies compared dawn simulation with an attentional control. There was some evidence that dawn simulation improved symptoms of depression but it was not clinically important and was not supported by other outcomes including the major depression with a seasonal pattern subscale. Similarly, there was no evidence of superiority of dawn simulation over regular light therapy. Evidence from the important outcomes and overall quality of evidence are presented in Table 97. The full evidence profiles and associated forest plots can be found in Appendix 16c and Appendix 19c, respectively.

Table 97

Summary evidence profile for dawn simulation studies.

Prevention of future episodes using light therapy

One study compared bight light therapy with a control treatment and with no treatment as relapse prevention in people who had a history of depression with a seasonal pattern but had not yet developed symptoms. This showed that those receiving light therapy were less likely to develop symptoms of depression compared with those receiving no treatment. However, those using the infrared light visor were less likely to develop symptoms of depression than those using the bright white light visor. Neither finding was clinically important. Evidence from the important outcomes and overall quality of evidence are presented in Table 98. The full evidence profiles and associated forest plots can be found in Appendix 16c and Appendix 19c, respectively.

Table 98

Summary evidence profile for relapse prevention using bright light.

11.6.4. Other therapies for depression with a seasonal pattern

Studies considered¹⁷⁴

In total, 14 trials of interventions other than bright light were found, mostly of anti-depressants, of which five met inclusion criteria for a review of acute-phase treatment, one for a review of continuation treatment in people who had responded to open-label treatment, and three (published in the same paper) for a review of prevention in people with a history of depression with a seasonal pattern. Summary study characteristics of the included studies are presented in Table 99, with full details in Appendix 17c, which also includes details of excluded studies.

Table 99

Summary study characteristics for interventions other than bright light for major depression with a seasonal pattern.

Clinical evidence

Acute-phase treatments

The data for acute-phase treatment comparing antidepressants with placebo were largely inconclusive, although on one outcome (response) there appeared to be little difference. Acceptability and tolerability data were inconclusive. There was no evidence to suggest a difference between moclobemide and fluoxetine, which was the only head-to-head evidence available. There was some evidence to suggest that high ion density was more effective than low ion density, although there was only one study. Evidence from the important outcomes and overall quality of evidence are presented in Table 100. The full evidence profiles and associated forest plots can be found in Appendix 16c and Appendix 19c, respectively.

Table 100

Summary evidence profile for acute-phase treatments (not light therapy) for major depression with a seasonal pattern.

Continuation treatment and prevention of future episodes

One small study compared the β-blocker, propanolol, with placebo for people who had responded to previous open treatment. This showed that symptoms of depression in those continuing treatment remained lower compared with those switched to placebo. Another three trials compared bupropion with placebo to prevent episodes in people with a history of depression. Treatment started before the onset of winter and continued until early spring. There was a clinically important reduction in the number of recurrences among those taking bupropion compared with the rate in those taking placebo. Evidence from the important outcomes and overall quality of evidence are presented in Table 101. The full evidence profiles and associated forest plots can be found in Appendix 16c and Appendix 19c, respectively.

Table 101

Summary evidence profile of continuation treatment and prevention of future episodes for people with major depression with a seasonal pattern.

11.6.5. From evidence to recommendations

The evidence for light therapy for major depression with a seasonal pattern is poorly developed, with many trials comparing different elements of treatment, including time of day, level of light and length of treatment. There is little evidence for the efficacy of bright light in the treatment of major depression with a seasonal pattern compared with placebo treatment.

The evidence for other treatments is sparse. Evidence is lacking that antidepressants are effective once symptoms have begun, but they may be worthwhile as prophylactics. For depression with a seasonal pattern practitioners should follow the guidance for depression elsewhere in this guideline.

11.6.6. Recommendations

11.6.6.1.

See recommendation 11.3.2.1

11.6.6.2.

Advise people with winter depression that follows a seasonal pattern and who wish to try light therapy in preference to antidepressant or psychological treatment that the evidence for the efficacy of light therapy is uncertain.

11.6.7. Research recommendations

11.6.7.1.

The efficacy of light therapy compared with antidepressants for mild to moderate depression with a seasonal pattern

How effective is light therapy compared with antidepressants for mild to moderate depression with a seasonal pattern?

11.7.1. Low-dose versus high-dose TCAs

**There is controversy over whether the existing recommended dosages for TCAs (100 mg/day, Bollini et al., 1999) are too high. Some GPs are criticised for prescribing at doses that are too low, and evidence for dosing levels has not been established (Furukawa et al., 2002a). This review compares the efficacy and tolerability of low and high doses of TCAs. Low doses were those where the mean dose achieved was less than the equivalent of 100 mg of amitriptyline.

11.7.2. Studies considered for review^175,176

The GDG used an existing review (Furukawa et al., 2002a) as the basis for this review. The Furukawa and colleagues’ (2002a) review included 38 studies of which 33 did not meet the inclusion criteria set by the GDG, mainly because of inadequate diagnosis of depression. Therefore, five trials (Burch1988, Danish1999, Rouillon1994, Simpson1988, WHO1986) are included in this review providing data from up to 222 participants.

All included studies were published between 1988 and 1999 and were between 4 and 8 weeks’ long (mean = 6 weeks). One study was of inpatients and two of out-patients, with none in primary care. Patients in one study were from mixed sources (Danish1999). It was not possible to discern the setting in WHO1986. No study included all elderly participants or those whose depression has atypical features. Study inclusion criteria ensured a minimum HRSD score at baseline of between 16 and 22 or a MADRS score of 15.

Data were available to compare low doses with high doses of clomipramine, amitriptyline, trimipramine and imipramine. Data were also available to compare low-dose clomipramine with placebo.

Mean low dose was 60.8 mg (total range 25 mg to 75 mg) and mean high dose was 161.9 mg (total range 75 mg to 200 mg) (low-dose versus high-dose studies).

11.7.3. Clinical evidence statements¹⁷⁷

11.7.4. Clinical summary

There is no clinically important difference on achieving response between low-dose TCAs (mean dose = 60.8 mg) and therapeutic dose TCAs (mean dose = 161.9 mg). Of the four studies that compared low-dose TCAs with high-dose TCAs, two reported completer data only. Patients receiving a low-dose TCA were less likely to leave treatment early due to side effects.**

11.7.5. From evidence to recommendations

This review was not updated by the GDG and the recommendation to maintain a low-dose TCA in people whose depression had responded was retained. However, the recommendation to monitor outcomes and increase dose depending on efficacy and side effects was removed since the points made are adequately covered by other recommendations in the guideline.

11.7.6. Recommendation

11.7.6.1.

People who start on low-dose tricyclic antidepressants and who have a clear clinical response can be maintained on that dose with careful monitoring¹⁷⁸.

11.8.1. Introduction

There can be confusion over the use of the terms ‘addiction’, ‘psychological dependence’ and ‘physical dependence’ when referring to drugs. This has been associated with concern in the mind of the public about whether antidepressants (and indeed other psychotropic drugs) may be addictive. The DSM–IV (APA, 1994) definition of ‘substance dependence’ consists of a combination of psychological, physiological and behavioural effects that together comprise what is commonly called addiction. The diagnosis of substance dependence/addiction requires at least three of the following:

1. tolerance (marked increase in amount; marked decrease in effect)
2. characteristic ‘withdrawal’ symptoms or substance taken to relieve withdrawal
3. substance taken in larger amount and for longer period than intended
4. persistent desire or repeated unsuccessful attempt to quit
5. much time/activity taken to obtain, use and recover from the substance
6. important social, occupational, or recreational activities given up or reduced
7. use continues despite knowledge of adverse consequences (for example, failure to fulfill role obligation, using when physically hazardous).

Physical dependence refers to the first two features (tolerance to the effect and ‘withdrawal’ symptoms) and substance dependence/addiction can be with or without physical dependence. There is no evidence that antidepressants cause psychological dependence or adverse behavioural and functional effects in the sense defined by criteria 3 to 7 above, and therefore antidepressants are not ‘addictive’ in the accepted sense of the word used to describe dependence on drugs like alcohol or opioids. There is also no good evidence to support tolerance to the therapeutic effect of antidepressants (Zimmerman & Thongy, 2007) and therefore the debate about whether or not antidepressants cause physical dependence centres on the symptoms some people experience when stopping antidepressants. It is important to understand the nature of the phenomenon and its implications for people with depression who have antidepressant treatment. In this guideline these are described as ‘discontinuation symptoms’, which is a term that makes no assumption about their status.

Discontinuation symptoms can be broadly divided into six groups; affective (for example, irritability), gastrointestinal (for example, nausea), neuromotor (for example, ataxia), vasomotor (for example, sweating), neurosensory (for example, paraesthesia), and other neurological (for example, dreaming; Delgrado, 2006). They may be new or hard to distinguish from some of the original symptoms of the underlying illness. By definition they must not be attributable to other causes. They are experienced by at least a third of patients (Lejoyeux et al., 1996; MHRA, 2004) and are seen to some extent with all antidepressants (Taylor et al., 2006). Of the commonly used antidepressants, the risk of discontinuation symptoms seems to be greatest with paroxetine, venlafaxine and amitriptyline (Taylor et al., 2006). There have been prospective studies, including some RCTs and quasi-randomised trials, which have examined the effect of discontinuation in people taking paroxetine and other anti-depressants. These studies suggest an increase in discontinuation symptoms in those taking paroxetine compared with escitalopram (Baldwin et al., 2006), fluoxetine (Rosenbaum et al., 1998; Bogetto et al., 2002; Hindmarch et al., 2000; Judge et al., 2002; Michelson et al., 2000), sertraline (Hindmarch et al., 2000; Michelson et al., 2000), and citalopram (Hindmarch et al., 2000). In addition two RCTs measuring discontinuation symptoms when stopping antidepressants after 8 weeks of treatment found that these were more common with venlafaxine than escitalopram (Montgomery et al., 2004) and moderate and severe symptoms were more common with venlafaxine compared with sertraline (Sir et al., 2005).

The onset is usually within 5 days of stopping treatment, or occasionally during taper or after missed doses (Rosenbaum et al., 1998; Michelson et al., 2000). This is influenced by a number of factors, which may include a drug’s half-life. Symptoms can vary in form and intensity and occur in any combination. They are usually mild and self-limiting, but can be severe and prolonged, particularly if withdrawal is abrupt. Some symptoms are more likely with individual drugs, for example dizziness and electric shock-like sensations with SSRIs, and sweating and headache with TCAs (Lejoyeux et al., 1996; Haddad, 2001).

11.8.2. Factors affecting the development of discontinuation symptoms

**Although anyone can experience discontinuation symptoms, the risk is increased in those prescribed short half-life drugs (Rosenbaum et al., 1998), such as paroxetine and venlafaxine (Fava et al., 1997; Hindmarch et al., 2000; MHRA, 2004). They can also occur in patients who do not take their medication regularly. Two-thirds of patients prescribed antidepressants skip a few doses from time to time (Meijer et al., 2001). The risk is also increased in those who have been taking antidepressants for 8 weeks or longer (Haddad, 2001); those who developed anxiety symptoms at the start of antidepressant treatment (particularly with SSRIs); those receiving other centrally acting medications (for example, antihypertensives, antihistamines, antipsychotics); children and adolescents; and those who have experienced discontinuation symptoms before (Lejoyeux & Ades, 1997; Haddad, 2001).

Discontinuation symptoms may also be more common in those who relapse on stopping antidepressants (Zajecka et al., 1998; Markowitz et al., 2000).

11.8.3. Clinical relevance

The symptoms of a discontinuation reaction may be mistaken for a relapse of illness or the emergence of a new physical illness (Haddad, 2001) leading to unnecessary investigations or reintroduction of the antidepressant. Symptoms may be severe enough to interfere with daily functioning. Another point of clinical relevance is that patients who experience discontinuation symptoms may assume that this means that antidepressants are addictive and not wish to accept further treatment. It is very important to counsel patients before, during and after antidepressant treatment about the nature of this syndrome.**

11.8.4. How to avoid discontinuation symptoms

Although it is generally advised that antidepressants (except fluoxetine) should be discontinued over a period of at least 4 weeks, preliminary data suggest that it may be the half-life of the antidepressant rather than the rate of taper that ultimately influences the risk of discontinuation symptoms (Tint et al., 2008).

When switching from one antidepressant to another with a similar pharmacological profile, the risk of discontinuation symptoms may be reduced by completing the switch as quickly as possible (a few days at most). A different approach may be required at the end of treatment where a slower taper is likely to be beneficial.

**The half-life of the drug should be taken into account. The end of the taper may need to be slower as symptoms may not appear until the reduction in the total daily dosage of the antidepressant is substantial. Patients receiving MAOIs may need dosage to be tapered over a longer period. Tranylcypromine may be particularly difficult to stop. It is not clear if the need for slow discontinuation of MAOIs, and particularly tranylcypromine, is due to the discontinuation syndrome or the loss of other neurochemical effects of these drugs. Since it is not possible to disentangle these phenomena, the clinical advice is that patients on MAOIs and those at-risk patients need a slower taper (Haddad, 2001).**

Many patients experience discontinuation symptoms despite a slow taper. For these patients, the option of abrupt withdrawal should be discussed. Some may prefer a short period of intense symptoms over a prolonged period of milder symptoms.

11.8.5. How to treat

**There are no systematic randomised studies in this area. Treatment is pragmatic. If symptoms are mild, reassure the patient that these symptoms are not uncommon after discontinuing an antidepressant and that they will pass in a few days. If symptoms are severe, reintroduce the original antidepressant (or another with a longer half-life from the same class) and taper gradually while monitoring for symptoms (Haddad, 2001; Lejoyeux & Ades, 1997).**

11.8.6. From evidence to recommendations

Since the previous guideline, the evidence base for discontinuation symptoms with antidepressants is largely unchanged. Practitioners should ensure that they discuss the issue fully with all patients, and consider prescribing antidepressants that are associated with fewer discontinuation symptoms (for example, fluoxetine), particularly for patients who have had previous experience of these. The previous recommendations are therefore retained, but rewritten to fit the updated NICE style.

11.8.7. Clinical practice recommendations

11.8.7.1.

When prescribing antidepressants, explore any concerns the person with depression has about taking medication, explain fully the reasons for prescribing, and provide information about taking antidepressants, including:

• the gradual development of the full antidepressant effect
• the importance of taking medication as prescribed and the need to continue treatment after remission
• potential side effects
• the potential for interactions with other medications
• the risk and nature of discontinuation symptoms with all antidepressants, particularly with drugs with a shorter half-life (such as paroxetine and venlafaxine), and how these symptoms can be minimised
• the fact that addiction does not occur with antidepressants.
  Offer written information appropriate to the person’s needs.

11.8.7.2.

Advise people with depression who are taking antidepressants that discontinuation symptoms¹⁷⁹ may occur on stopping, missing doses or, occasionally, on reducing the dose of the drug. Explain that symptoms are usually mild and self-limiting over about 1 week, but can be severe, particularly if the drug is stopped abruptly.

11.8.7.3.

When stopping an antidepressant, gradually reduce the dose, normally over a 4-week period, although some people may require longer periods, particularly with drugs with a shorter half-life (such as paroxetine and venlafaxine). This is not required with fluoxetine because of its long half-life:

11.8.7.4.

Inform the person that they should seek advice from their practitioner if they experience significant discontinuation symptoms. If discontinuation symptoms occur:

• monitor symptoms and reassure the person if symptoms are mild
• consider reintroducing the original antidepressant at the dose that was effective (or another antidepressant with a longer half-life from the same class) if symptoms are severe, and reduce the dose gradually while monitoring symptoms.

11.9.1. Introduction

**Consistent associations between depression and cardiovascular morbidity and mortality have been identified (Glassman & Shapiro, 1998). Depression is a significant independent risk factor for both first myocardial infarction and cardiovascular mortality with an adjusted relative risk in the range of 1.5 to 2 (Ford et al., 1998). In patients with ischaemic heart disease, depression has been found to be associated with a three- to four-fold increase in cardiovascular morbidity and mortality (Carney et al., 1997). The prevalence of depression in patients with coronary heart disease is approximately 20% (Glassman et al., 2002).

In view of the above associations and factors it is important to use antidepressant drugs that either reduce or do not increase the cardiovascular risk of the condition itself and to establish a safe and effective treatment strategy for depressed patients with heart disease. There is evidence that adequate treatment of depression appears either to lower (Avery & Winokur, 1976) or not to change (Pratt et al., 1996) the risk of heart disease. However, two large-scale follow-up studies have shown an increase in myocardial infarction in users of antidepressants with an average odds ratio of 5.8 (Penttinen & Valonen, 1996; Thorogood et al., 1992). The antidepressants used in these studies were predominately TCAs. A similar association has been identified in the UK for dothiepin/dosulepin (Hippisley-Cox et al., 2001).

However, these studies do not distinguish between the effects of drugs and the condition itself. Thus it is necessary to look at the effects of antidepressants on cardiovascular function and what trials are available (Roose, 2003).

11.9.2. Tricyclic antidepressants

Sinus tachycardia, postural hypotension and episodic hypertension are side effects frequently observed. Electrocardiogram (ECG) changes are frequent, such as lengthening of the QT, PR and QRS intervals relating to alterations in atrioventricular conduction and repolarisation (Roose & Glassman, 1989). These effects are due to the wide-ranging pharmacological actions of TCAs that are not correlated with recognised mechanisms of antidepressant action. In healthy patients such changes may be asymptomatic or clinically unimportant, but in those with heart disease they may lead to significant morbidity and mortality (Glassman et al., 1993). For example, prolonged increased heart rate (mean 11%, Roose & Glassman, 1989) could have a major impact in terms of cardiac work (Roose, 2003).

In patients with left ventricular impairment on TCAs, orthostatic hypotension is three to seven times more common and potentially clinically harmful (Glassman et al., 1993). The TCA induced prolongation of conduction may be clinically unimportant in healthy patients, but can lead to complications in those with conduction disease, in particular bundle branch block, and these can be severe in 20% of subjects (Roose et al., 1987). TCAs may be regarded as Class I arrhythmic drugs. Evidence suggests that this class of drug is associated with an increase in mortality in post-infarction patients and in patients with a broader range of ischaemic disease, probably because they turn out to be arrhythmogenic when cardiac tissue becomes anoxic. Overdose of TCAs or elevated plasma levels as a result of interactions with other drugs, liver disease and age is associated with serious hypotension and atrial and ventricular arrhythmias may arise even to the extent of complete atrioventricular block, which in a number of cases may be fatal (deaths from TCAs represent 20% of overdose deaths; Shah et al., 2001).

11.9.3. Selective serotonin reuptake inhibitors

Depression in untreated populations has been demonstrated to increase cardiovascular morbidity and mortality. SSRIs appear to reduce that risk, since two studies have reported no difference in cardiovascular risk between SSRI-treated depressed patients and non-treated non-depressed controls (Cohen et al., 2000; Meier et al., 2001). Sauer and colleagues (2001) compared the rate of myocardial infarction (MI) in patients on an SSRI with those on no antidepressants. The SSRI-treated patients had a significantly lower rate of MI than did the non SSRI-treated patients. Multiple studies (Roose, 2001) reveal no clinically significant effects of SSRIs on heart rate, cardiac conduction or blood pressure (see further details below). Studies of depressed patients with and without ischaemic heart disease (IHD) have documented increased platelet activation and aggregation, which potentially contributes to thrombus formation (Musselman et al., 1998). Treatment with SSRIs normalises elevated indices of platelet activation and aggregation seen in non-treated patients with depression and IHD. There is evidence that this effect occurs at relatively low doses and before the antidepressant effect (Pollock et al., 2000). However, the effects on platelet serotonin are not always advantageous: SSRIs increase the probability of having a serious gastrointestinal bleed, particularly in the very old (Walraven et al., 2001).

11.9.4. Other drugs

11.9.5. Recommendation

11.9.5.1.

See recommendation 10.14.1.3.

11.10.1. Introduction

**The majority of patients with depression have at least episodic suicidal ideation often linked to general negativity and hopelessness. Two-thirds of people who attempt suicide are experiencing depression, and suicide is the main cause of the increased mortality of depression and is commonest in those with comorbid physical and mental illness. Suicidal behaviour also occurs with milder forms of depression. In a meta-analysis of 36 studies the lifetime prevalence of suicide has been reported to be 4% in hospitalised depressed patients, rising to 8.6% if hospitalised for suicidality. In mixed inpatient/outpatient populations the lifetime prevalence is 2.2% compared with less than 0.5% in the non-affectively ill population (Bostwick & Pankratz, 2000). Harris and Barraclough (1997) found a suicide risk of 12 times that expected in a cohort of patients with dysthymia (DSM–III) (APA, 1980). Therefore, the effective recognition and treatment of depression should lead to a fall in the overall suicide rate.

11.10.2. Suicidality and antidepressants

There is evidence for a small but significant increase in the presence of suicidal thoughts in the early stages of antidepressant treatment (Jick et al., 2004). However this must be put against recent data showing that the risk of clinically important suicidal behaviour is highest in the month before starting antidepressants and declines thereafter (Simon et al., 2006). The highest rates of suicidal behaviour were seen in patients treated by psychiatrists but the same pattern was also seen with psychological treatments and in primary care (Simon & Savarino, 2007). No temporal pattern of completed suicide was found in the 6 months after starting an antidepressant (Simon et al., 2006). No increase in suicide/suicidal thoughts or attempts was seen with SSRIs compared with other antidepressants (Jick et al., 2004; Simon et al., 2006).

It is therefore not clear from these naturalistic data to what extent suicidal thoughts or behaviour can be attributable to a direct result of taking an antidepressant (the effect was seen with all classes of antidepressant) as opposed to the timing of when help was sought. Two meta-analyses of RCTs (Fergusson et al., 2005; Gunnell et al., 2005) with 702 and 477 studies respectively and a large nested case-control study comparing new prescriptions of SSRIs and TCAs (Martinez et al., 2005) found no evidence of an increase in completed suicide with SSRIs but possible evidence of increased suicidal/self-harming behaviour with SSRIs compared with placebo (NNH 684 and 754 in the two meta-analyses). There was no overall difference between SSRIs and TCAs (Fergusson et al., 2005; Martinez et al., 2005) but Martinez and colleagues (2005) found some evidence for increased self-harming behaviour with SSRIs compared with TCAs in those under 19 years. A review by Möller and colleagues (2008) concluded that all antidepressants carry a small risk of inducing suicidal thoughts and suicide attempts in age groups below 25 years, the risk reducing further at the age of about 30 to 40 years.

There may be a delay in noticeable improvement after starting antidepressants, and, just after initiation of treatment, mood remains low with prominent feelings of guilt and hopelessness, but energy and motivation can increase and may be related to the increased suicidal thoughts. A similar situation can arise with patients who develop akathisia or increased anxiety due to a direct effect of some SSRIs and related drugs and it has been hypothesised that this may increase the propensity to suicidal ideation and suicidal behaviour (Healey, 2003). Careful monitoring is therefore indicated when treatment is initiated with an antidepressant. Patients should be monitored regardless of the apparent severity of their depression.

A meta-analysis of observational studies (Barbui et al., 2009) found that compared with depressed people who did not take antidepressants, adolescents receiving SSRIs had a significantly higher risk of suicide attempts and completed suicide. In contrast adults, especially older adults, had a significantly lower risk of suicide attempts and completed suicide. Ecological data has failed to find any link between SSRI use and higher completed suicide rates (Gibbons et al., 2005; Hall & Lucke, 2006), in fact it has been suggested that the overall reduction in suicide rate may be partly due to more effective treatment of depression with newer antidepressants. In particular, it has been argued that the significant reductions in suicide rates in Sweden, Hungary, the US and Australia have been due to treatment with these drugs (Isacsson et al., 1997; Hall et al., 2003). However, a number of other factors may account for this trend including changing socioeconomic circumstances, and demonstrating a causal link between increased antidepressant prescription and falling suicide rates is not straightforward and has not been conclusively established (Gunnell & Ashby, 2004).

The use of antidepressants in the treatment of depression is also not without risk not least because of their toxicity in overdose. Antidepressants were involved in 18% of deaths from drug poisoning between 1993 and 2002 (Morgan et al., 2004), with TCAs, which are cardiotoxic in overdose (see Section 8.2.9), accounting for 89% of these. This is equivalent to 30.1 deaths per million prescriptions. Dothiepin/dosulepin alone accounted for 48.5 deaths per million prescriptions (Morgan et al., 2004). By contrast, over the same period, SSRIs accounted for around 6% of deaths by suicide, and other antidepressants, including venlafaxine, around 3%. This is equivalent to 1 and 5.2 deaths per million prescriptions respectively (Morgan et al., 2004). Venlafaxine alone accounted for 8.5 deaths per million prescriptions. Morgan and colleagues (2004) showed an overall reduction in mortality rates over the time period studied, with a fall in rates related to TCAs, little change for SSRIs, but an increase for other antidepressants largely due to venlafaxine. These data are based on analyses of coroners’ records for England and Wales, and prescription data for drugs dispensed in England (regardless of the prescription’s country of origin). They may be subject to bias because indication is not recorded on prescriptions. Some antidepressants are licensed for conditions such as obsessive-compulsive disorder and post-traumatic stress disorder in addition to depression. Also, coroners record antidepressant information voluntarily and only if they consider the antidepressant contributed to the cause of death (Morgan et al., 2004). Interpretation of these data is complicated by the possibility of differential prescribing, that is patients at high risk of suicide may have been prescribed different drugs from those at low risk.** The MHRA (2006a and b) concluded that the increased rate seen with venlafaxine was partly, but not wholly, attributable to patient characteristics.

11.10.3. From evidence to recommendations

There is a small risk of inducing suicidal ideation in younger people starting anti-depressants. Although the most recent data suggests the cut-off for this is around 25 years old, previous advice from the MHRA suggests the cut-off should be around 30. Practitioners should seek strategies to reduce risk as far as possible for people who are at increased risk of suicide, including prescribing drugs with relatively low toxicity and prescribing small amounts of drugs. They should refer people at high risk to specialist mental health services. The recommendations in this section are unchanged from the previous guideline, but have been reworded to fit current NICE house style and to fit with new recommendations developed for the updated guideline.

11.10.4. Recommendations

11.10.4.1.

A person with depression started on antidepressants who is considered to present an increased suicide risk or is younger than 30 years (because of the potential increased prevalence of suicidal thoughts in the early stages of antidepressant treatment for this group) should normally be seen after 1 week and frequently thereafter as appropriate until the risk is no longer considered clinically important¹⁸¹.

11.10.4.2.

See recommendation 5.2.13.15.

11.10.4.3.

See recommendation 5.2.13.13.

11.10.4.4.

Take into account toxicity in overdose when choosing an antidepressant for people at significant risk of suicide. Be aware that:

• compared with other equally effective antidepressants recommended for routine use in primary care, venlafaxine is associated with a greater risk of death from overdose
• tricyclic antidepressants (TCAs), except for lofepramine, are associated with the greatest risk in overdose.